Predicting Surgical Outcome in Drug-Resistant Epilepsy by Combining Interictal Biomarkers within a Machine Learning Framework

Predicting Surgical Outcome in Drug-Resistant Epilepsy by Combining Interictal Biomarkers within a Machine Learning Framework | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Predicting Surgical Outcome in Drug-Resistant Epilepsy by Combining Interictal Biomarkers within a Machine Learning Framework Hmayag Partamian, Saeed Jahromi, M Scott Perry, Eleonora Tamilia, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8682213/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 26 Mar, 2026 Read the published version in Scientific Reports → Version 1 posted 11 You are reading this latest preprint version Abstract Delineating the epileptogenic zone (EZ) is essential for achieving seizure freedom in drug-resistant epilepsy (DRE). Conventionally, seizure onset derived from ictal intracranial EEG (iEEG) approximates the EZ, but acquiring ictal data can be challenging. Interictal iEEG abnormalities offer abundant, easily acquired, non-seizure-dependent markers of the epileptogenic tissue; however, these biomarkers offer limited specificity. Here, we propose a machine-learning framework that integrates interictal spike and ripple features to automatically delineate the EZ and predict outcome with improved performance compared to individual biomarkers. We retrospectively analyzed iEEG data from 62 children with DRE ([34 good (Engel 1) outcomes] undergoing neurosurgery, automatically detected spikes and ripples, and computed temporal, spectral, and spatial features for each channel. We trained Random Forest classifiers to predict the EZ using combinations of these features. The predicted EZ derived from spike-based and combined spike-ripple feature sets outperformed those from individual biomarkers in defining the EZ, with an area under the receiver operating characteristic curve of 0.9 and 74% spatial overlap with resection. Although most individual features and classifiers predicted the outcome, the combined feature model performed best (i.e., sensitivity 88%, specificity 68%, and accuracy 79%). Our findings demonstrate that integrating multimodal interictal features improves the EZ delineation, providing valuable prognostic insights for epilepsy surgery. Health sciences/Biomarkers Biological sciences/Computational biology and bioinformatics Health sciences/Neurology Biological sciences/Neuroscience interictal epilepsy biomarkers epilepsy surgery combined interictal features supervised machine learning Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Precise delineation of the epileptogenic zone (EZ) remains one of the most significant challenges for surgery planning and achieving seizure freedom post-surgery in patients with drug-resistant epilepsy (DRE). 1 , 2 This hypothetical zone is often approximated by the seizure onset zone (SOZ), which is identified by manual review of ictal activity captured by intracranial EEG (iEEG) to detect the brain region(s) that initiate clinical seizures. 1 Yet, increasing evidence suggests that interictal electrophysiological abnormalities observed in iEEG recordings (i.e., brief pathological events occurring between seizures) may also approximate the EZ. 3,4 Importantly, identifying such biomarkers from readily available interictal data offers a significant advantage by eliminating the need to wait for a seizure to occur. 5 , 6 Among the proposed interictal epilepsy biomarkers, spikes and high-frequency oscillations (HFOs) are the most extensively studied. 3 , 4 , 7 Interictal spikes, brief high-amplitude deflections typically occurring in the 1–70 Hz frequency band, are routinely used in clinical practice as markers of the irritative zone. 6 , 8 , 9 HFOs, which are considered more specific biomarkers of the EZ compared to spikes 10 , 11 , are commonly classified into ripples (80–250 Hz) and fast ripples (250–500 Hz). 12 – 14 Fast ripples, although closely linked to the EZ 15–17 , are less frequently observed in patients with DRE, and their low sensitivity may limit their value for the EZ localization. 18 , 19 On the other hand, ripples are present in most patients with DRE but are also observed in physiological brain regions. 6 , 10 , 20 , 21 Thus, the presurgical value of interictal biomarkers, such as the spikes and HFOs, as biomarkers of the EZ remains largely debated since the cortical areas that generate them often extend beyond the EZ into brain regions that do not need to be resected or ablated for the patient to become seizure-free. 11 , 20 , 22 , 23 Because of these limitations, temporal (i.e., timing and temporal occurrence), spectral (i.e., frequency content), and spatial (i.e., distribution across brain regions) features of spikes and ripples have been examined as potential indicators of epileptogenicity. In the temporal domain, the spike and ripple rates have been linked to the underlying brain tissue’s epileptogenicity. 22 , 24 , 25 Yet, spike-generating regions correspond to the broader irritative zone, which can extend beyond the EZ and overlap with physiological brain regions. 26 , 27 Moreover, not all ripples are pathological, as physiological ripples also occur, particularly in brain regions such as the hippocampus, occipital cortex, and paracentral areas. 28 – 30 Consequently, spike and ripple rates alone are unreliable predictors of the EZ. Recent studies have shown that the temporal co-occurrence of spikes and ripples may predict the EZ more effectively, or with comparable efficacy, compared to spikes or ripples alone. 22 , 25 , 31 This improved performance likely reflects the co-occurrence of spikes and ripples as a more pathologic pattern of neuronal synchronization than either event in isolation. 32 Yet, the interpretation of ripples remains challenging due to their susceptibility to contamination by artifacts arising from the Gibbs phenomenon, particularly in recordings containing epileptiform spikes. 33 , 34 In the spectral domain, frequency-based features have been shown to identify abnormal oscillatory activity associated with epileptogenic networks. 35 – 37 However, they are limited by their sensitivity to sensor noise and biological artifacts. 38 , 39 Spatial analyses have demonstrated that spikes and ripples often originate in specific brain regions and propagate to adjacent and distant areas within a few tens of milliseconds. 10 , 11 , 40 , 41 Recent studies have also shown that the onset generator of spikes and ripples (i.e., regions where discharges show the earliest peaks) are correlated with the EZ, 10,40,42,43 often outperforming the clinically defined SOZ in predicting the surgical outcome. 11 , 40 , 44 However, the spatial distribution of spikes and ripples fluctuates over time, shifting among multiple regions rather than arising from a single consistent source. 45 , 46 Therefore, no single biomarker fully defines the EZ; rather, each reflects only a partial and variable aspect of epileptogenicity, highlighting the need for complementary approaches. It is reasonable to presume that the integration of temporal, spectral, and spatial features of interictal biomarkers will yield a more comprehensive and precise delineation of the EZ. 47–49 Yet, the practical implementation of this approach is technically challenging. iEEG recordings are often performed with several dozens of contacts sampled at high rates (> 1 kHz) and collected over several hours. As a result, they generate a massive amount of data, making the process time-consuming, labor-intensive, and prone to bias and error. This underscores the need to develop novel automated approaches that integrate the complementary strengths of these biomarkers and features into a unified framework, enabling a more objective, efficient, and accurate characterization of the EZ. Machine learning (ML) offers a robust framework for automating the analysis of large-scale data and integrating heterogeneous, high-dimensional features that are difficult to analyze manually or with conventional statistical methods. 50 , 51 Previous studies have incorporated temporal, spectral, and spatial features using ML to delineate the EZ in patients with DRE. 37 , 52 – 59 However, few studies have further compared the performance of individual features with multi-feature ML-based frameworks’ findings. 53 , 60 This comparison is critical, as the epileptogenic brain regions identified by individual spike and ripple features may differ across patients, or map incompletely to the EZ or extend beyond it. These limitations highlight the need for methods that systematically integrate heterogeneous biomarkers for EZ prediction. ML-based frameworks are well-suited for this task, as they can fuse multiple feature types, thereby improving accuracy, robustness, and interpretability in EZ mapping and surgical outcome prediction. In this study, we propose an ML-based multimodal framework that integrates diverse interictal spike and ripple temporal, spectral, and spatial features to automatically identify the EZ and predict postsurgical outcome in patients with DRE. We hypothesize that the proposed framework outperforms single-feature analyses in delineating the EZ and predicting the outcome by capturing nonlinear and cross-domain relationships among epileptogenic biomarkers. To evaluate this framework, automatic spike and ripple detectors were applied to iEEG recordings from children and young adults with DRE to extract multidomain electrode-level features. These features were then used to train Random Forest classifiers, with different combinations of features as predictors and resection as the target, to define the predicted EZ (PEZ). The PEZs derived from individual spike and ripple features and from the proposed feature-combining approach were analyzed to evaluate their ability to predict the EZ. We then assessed whether resection of each PEZ could predict surgical outcome. By combining multimodal interictal features, this study establishes a data-driven framework for more accurate localization of the EZ and prediction of surgical outcome for patients with DRE. Results Patient cohort We retrospectively analyzed interictal iEEG data from 62 children and young adults (29 females) with focal onset DRE, having either a good (34 patients, Engel I) or poor (28 patients, Engel \(\:>\) I) surgical outcome following epilepsy surgery [median follow-up: 2.5 years (2–5)], who satisfied the study’s inclusion and exclusion criteria (see Materials and Methods ). Patient demographics are summarized in Table 1 . The patients had a median age at surgery of 12 years (range: 8-15.5) with a median seizure onset age of 4.5 years (2–9). Our dataset included patients with three implantation types: (i) subdural electrodes [electrocorticography (ECoG)] (13 patients; eight with good outcome); (ii) stereotactic EEG (sEEG) implantation (36 patients; 20 with good outcome); and (iii) both subdural and depth electrodes (13 patients; 6 with good outcome). The median number of implanted contacts was 124 (102–151). Seventeen patients (27.4%) had non-lesional MRI, 34 patients (54.8%) had a malformation of cortical development, and 11 patients (17.8%) had acquired brain injury. Twenty-six patients (41.9%) had temporal lobe epilepsy. No significant differences in these characteristics were found between good and poor outcome patients (Table 1 ). Table 1 Patient demographic information categorized by post-surgical outcome. Characteristic Total Good outcome (Engel = I) Poor outcome (Engel > I) p Patients 62 34 28 - Gender 0.17 a Male 31 18 13 Female 31 16 15 Age at surgery , yr., median (IQR). 12 (8-15.5) 12 (7-15.5) 12 (8–15) 0.8 b Age at seizure onset , yr., median (IQR). 4.5 (2–9) 6 (2-10.25) 4 (2-6.5) 0.08 b Follow up , yr., median (IQR). 2.5 (2–5) 2.5 (2–5) 2.5 (1–5) 0.8 b Number of contacts , no., median (IQR) 124 (102–151) 112 (88–136) 138 (106–168) 0.09 b Type of implantations 0.73 a Subdural (ECoG) 13 8 5 sEEG 36 20 16 Subdural and depth 13 6 7 Implantation side 0.96 a L 40 22 18 R 16 9 7 L + R 6 3 3 Eplilepsy localization 0.16 a Temporal 26 17 9 Extra-temporal 36 17 19 MRI Findings 0.73 a NL 17 9 8 MCD 34 20 14 ACQ 11 5 6 ACQ = Acquired (i.e., stroke, neoplasm, and traumatic brain injury); ECoG = Electrocorticography; IQR = Inter-Quartile Range; L = Left; MCD = Malformation of Cortical Development (i.e., focal cortical dysplasia, polymicrogyria, gliosis, and tuberous sclerosis complex); NL = Non-Lesional; no. = Number; p = p -values; R = Right; sEEG = stereotactic electroencephalography; yr. = Year. a Chi-squared test; b Wilcoxon rank-sum test (two-tailed). Overall processing Pipeline The proposed framework automatically identifies the EZ from interictal iEEG data through automated detection of spikes and ripples, spike and ripple features extraction, and ML (Fig. 1 ). We initially preprocessed a 5-minute interictal iEEG segment to remove artifacts and noisy channels (Fig. 1 a). The data were then bandpass filtered into the spike (1–70 Hz) 40 and ripple (80–250 Hz) 44 frequency bands. Spikes and ripples were detected using previously validated fully automated algorithms. 10 , 62 From the detected spikes and ripples, we extracted a set of temporal, spectral, and spatial features. For the delineation of the EZ, we used as gold standard the resected volume of patients with good surgical outcome: channels inside the resection were labeled as EZ, and those outside as non-EZ. We first evaluated the individual discriminatory power of each feature for classifying the EZ vs. non-EZ channels. We then trained Random Forest classifiers to automatically identify the predicted EZ (PEZ). Finally, we assessed whether resecting the PEZ, which was mapped by each feature and the classifiers, could predict surgical outcome (Fig. 1 a). To characterize the epileptogenic properties of each channel, we derived a set of temporal, spectral and spatial features from the detected spikes and ripples (Fig. 1 b-f). Using in-house algorithms, we first identified propagation sequences of both spike and ripple events (Fig. 1 b-c). For each type of event, we determined the onset and spread channels and derived multiple quantitative features per channel, including propagation indices (Fig. 1 b-c), power (Fig. 1 d), spatial distribution (Fig. 1 e), and spike and ripple rates (Fig. 1 f). Finally, we computed the co-occurrence of spikes and ripples for each channel (see Materials and Methods ). Table 2 summarizes the features adopted in this study. The selected features were designed to capture complementary temporal, spectral, and spatial characteristics of interictal spikes and ripples, known to reflect epileptogenic tissue. Spike and ripple rates (i.e. SR and RR, respectively) and power (i.e. SP and RP, respectively) quantify local excitability and pathological high-frequency activity 63 , 64 , while spike and ripple propagation indices (i.e. SPI and RPI, respectively) characterize onset and spread channels. 65 The spatial distribution of interictal epileptiform events has been shown to reflect underlying epileptogenic organization and network spread, with events clustering around putative onset regions and dispersing along pathological propagation pathways. 48 We used several distance-based measures to quantify the spatial distribution of epileptiform events relative to the contact with the highest spike and ripple rate (i.e. ED-HSR and ED-HRR, respectively), the highest spike and ripple power (i.e. ED-HSP and ED-HRP, respectively), and the spike and ripple onset (i.e. ED-SO and ED-RO, respectively) contacts. In addition, we included the spike–ripple co-occurrence (SRC) to quantify pathological coupling between spikes and ripples, which reflects abnormal local network synchronization and has been strongly associated with epileptogenicity. 22 , 65 Together, these features integrate local and network-level information, providing a physiologically grounded and interpretable representation for automated delineation of the EZ. Table 2 Summary of spike and ripple features used for EZ characterization. Biomarker Feature Description Type Normalization Physiological Meaning Spike (1–70 Hz) SR Spike rate (min⁻¹) Temporal Min–Max Interictal spike excitability SP Spike power Spectral Min–Max Signal energy SPI Spike propagation index Temporal Max-Min Degree of spike spread across channels ED-HSR Euclidean distance from highest spike rate contact Spatial Max-Min Proximity to maximal spike-generating region ED-HSP Euclidean distance from highest spike power contact Spatial Max-Min Proximity to strongest spike source ED-SO Euclidean distance from spike onset contact Spatial Max-Min Distance from earliest spike onset Ripple (80–250 Hz) RR Ripple rate (min⁻¹) Temporal Min–Max High-frequency excitability RP Ripple power Spectral Min–Max Signal energy RPI Ripple propagation index Temporal Max-Min Spatial spread of ripple activity ED-HRR Euclidean distance from highest ripple rate contact Spatial Max-Min Proximity to dominant ripple-generating region ED-HRP Euclidean distance from highest ripple power contact Spatial Max-Min Proximity to strongest ripple source ED-RO Euclidean distance from ripple onset contact Spatial Max-Min Distance from earliest ripple onset Spike–Ripple SRC Spike–ripple co-occurrence rate Temporal Min–Max Pathological coupling of spikes and ripples We next coregistered the pre- and post-operative MRIs with the electrode implantation computerized tomography (CT) to map the locations of iEEG electrodes and the resected brain regions (Fig. 1 g). Pediatric epileptologists defined the SOZ through visual inspection of ictal iEEG data independently from this study. We finally trained Random Forest classifiers using different subsets of the extracted features as predictors: spike features alone (SF), ripple features alone (RF), and combined spike–ripple features (CF). Resected contacts were labeled as epileptogenic, and all remaining contacts were considered non-epileptogenic (Fig. 1 h). The contacts predicted as epileptogenic defined the PEZ. Training was performed only on good outcome patients, as their resections were assumed to contain the actual EZ, and model performance was evaluated using leave-one-patient-out cross-validation (LOPO-CV). We then quantified the PEZ with spatial measures: (i) focality ( F PEZ ); (ii) overlap with resection ( O RES ), and (iii) average distance from resection ( D RES ). 37 We defined the F PEZ as the average Euclidean distance (in mm) between the PEZ contacts. We defined the \(\:{O}_{RES}\:\) as the percentage of PEZ contacts within 10 mm of resection. We defined the D RES as the average Euclidean distance of the PEZ contacts to the resection. While training was performed using only good outcome patients, testing included both good and poor outcome patients. Finally, we tested whether the resection of the PEZ was associated with surgical outcome using the \(\:{O}_{RES}\:\) as predictor. Additional analyses compared the performance of individual features and classifiers trained on different feature subsets in predicting the EZ and outcome. We also assessed how feature selection and implantation type impacted the accuracy of EZ and outcome prediction (see Materials and Methods ). Predictive power of individual features To assess the relationship between the extracted features and the EZ, we compared the normalized feature values inside vs. outside resection in good and poor outcome patients, separately. In good outcome patients, we observed increased SR ( P < 0.001), SP ( P < 0.001), SPI ( P < 0.01), ED-HSR ( P < 0.0001), ED-HSP ( P < 0.0001), ED-SO ( P < 0.0001), RP ( P < 0.001), ED-HRR ( P < 0.01), ED-HRP ( P < 0.05), ED-RO ( P < 0.001), and SRC ( P < 0.05) inside (compared to outside) resection (Fig. 2 a). In poor outcome patients, we observed higher SR ( P < 0.05), SP ( P < 0.05), ED-HSP ( P < 0.05), ED-SO ( P < 0.01), RP ( P < 0.05), and ED-RO ( P < 0.05) inside (compared to outside) resection (Fig. 2 a). To assess the predictive power of each feature, we computed their performance in predicting the EZ in good outcome patients (Fig. 2 b). First, we evaluated feature importance using one-way ANOVA by ranking the proportion of variance (η²) explained by class labels (epileptogenic vs. non-epileptogenic) in decreasing order (see Materials and Methods ). The top-ranked features ED-SO, ED-HSP, SR, ED-HSR, SPI, and ED-RO exhibited strong discriminative power in differentiating between epileptogenic and non-epileptogenic contacts (η²>0.06) (Fig. 2 b). Next, we trained single-feature threshold-based classifiers and their performance in discriminating between epileptogenic and non-epileptogenic contacts using different performance metrics including sensitivity, specificity, precision, negative predictive value (NPV), accuracy, F1-measure, as well as the area under the receiver operating characteristic curve (AUC-ROC), precision–recall curve (AUC-PR), and \(\:{O}_{RES}\) (see Materials and Methods ). Among all features, SR, ED-HSR, ED-HSP, and ED-SO achieved the highest AUC-ROC and AUC-PR (> 74% and 66%, respectively) and exhibited > 60% overlap with resection. The resection of the PEZ, estimated from most single-feature classifiers, also predicted outcome, except for the SPI, RR, and RPI (Fig. 2 b). Predictive power of feature sets The performance of the Random Forest classifiers using different feature sets (i.e., SF, RF, CF) as predictors of the EZ was evaluated using the LOPO-CV, AUC-ROC, AUC-PR, and \(\:{O}_{RES}\) computed for individual features and feature sets (see Materials and Methods ). We found that SF and CF achieved higher AUC-ROC (0.9 and 0.89, respectively) compared to most individual features, except for the ED-SO (0.9). In terms of AUC-PR, SF, CF, and ED-SO showed higher values compared to all individual features (AUC-PR = 0.76, 0.78, and 0.77, respectively). Many individual features, however, showed higher AUC-ROC and AUC-PR compared to RF. Regarding overlap with resection, SF, RF, and CF all achieved \(\:{O}_{RES}\:\) > 74%, which was higher than the individual features ( \(\:{O}_{RES}\:\) ≤ 70%) (Fig. 3 a). Statistical pairwise comparisons between the performance measures of individual features and classifiers showed that SF and CF outperformed all individual features in terms of AUC-ROC and AUC-PR, except for the ED-HSR, ED-HSP, and ED-SO. RF showed higher performance than some individual features but did not exceed SF or CF. No differences were observed between SF and CF for the AUC-ROC and AUC-PR. Regarding the \(\:{O}_{RES}\) , SF and CF outperformed all individual features except the SRC, while no differences were observed among the SF, RF, and CF (Fig. 3 b). Effect of implantation type on EZ predictability To determine whether the implantation type (i.e., ECoG, sEEG, or both subdural and depth electrodes) influenced the PEZ's predictive power, we trained Random Forest classifiers separately for cohorts sharing the same implantation type. For good outcome patients, we estimated the PEZ using SF, RF, and CF and then calculated the AUC-ROC, AUC-PR, and \(\:{O}_{RES}\) (see Materials and Methods ). We found no differences between the AUC-ROC, AUC-PR, and \(\:{O}_{RES}\:\) of the PEZs with resection among the different implantation types for any of the feature sets (see Supplementary Note 2 ). Focality, proximity, and overlap of PEZ with resection To assess the concordance of PEZ with RES, we compared \(\:{O}_{RES}\) , \(\:{D}_{RES}\) , and \(\:{F}_{PEZ}\) across patients with good and poor outcomes for the three feature subsets. We found that the SF, RF, and CF had higher \(\:{O}_{RES}\) and lower \(\:{D}_{RES}\) in good compared to poor outcome patients ( P < 0.01, d \(\:\ge\:0.4\:\) ). No differences in \(\:{F}_{PEZ}\) were observed between good and poor outcome patients for any of the feature sets. Notably in good outcome patients, the SF had a median \(\:{O}_{RES}\) \(\:\approx\:\) 74 (56–95)%, RF had a median \(\:{O}_{RES}\) \(\:\approx\:\) 75 (50–85)%, and CF had a median \(\:{O}_{RES}\) \(\:\approx\:\) 74 (57–94)% (Fig. 4 a). EZ and Surgical Outcome Prediction Across Features and Classifiers To evaluate the PEZ as an interictal biomarker of the EZ, we examined its ability to predict surgical outcome (see Materials and Methods ). We found that all PEZ derived from SF, RF, and CF predicted outcome ( P < 0.01, Fisher Exact test), while the clinically defined SOZ did not predict outcome (Fig. 4 c). Additionally, the CF model predicted outcome with the highest sensitivity (88%), specificity (68%), precision (77%), NPV (83%), F1-score (82%), and accuracy (79%) (Fig. 4 c). We finally performed a pairwise McNemar test on the confusion matrices of outcome predictions of individual features, feature sets (i.e., SF, RF, CF), and the SOZ, to determine whether their outcome predictions were significantly different (Fig. 4 c). We found that SR was better than RPI and the SOZ, and ED-SO was better than the SOZ ( P < 0.05, McNemar test). SF showed better predictive power compared to the RPI and the SOZ ( P < 0.05, McNemar test). Finally, CF had higher predictive power of surgical outcome than all individual features except the SR, ED-HSR, ED-HSP, ED-SO, ED-HRP, SF, and RF ( P < 0.05, McNemar test). Representative cases To illustrate how the PEZ relates to the EZ, we report findings from four representative cases: two patients with good and two patients with poor surgical outcome. For these patients, we depict the PEZ overlayed on their MRIs and the resection volume (Fig. 5 ). For good outcome patient #6, the PEZ derived from the CF ( O RES = 71%) and SF ( O RES = 51%) showed greater overlap with the resection volume than that from the RF ( O RES = 41%) (Fig. 5 ). The PEZs from the CF and SF were also more focal ( \(\:{F}_{PEZ}\) =22 mm and 23 mm, respectively) and closer to the resection ( \(\:{D}_{RES}\) =12 mm and 13 mm, respectively) compared to the RF ( \(\:{F}_{PEZ}\) =32 mm, \(\:{D}_{RES}\) =19 mm). For good outcome patient #19, PEZs from all three classifiers (i.e., SF, RF, and CF) had high overlap with the resection ( \(\:{O}_{RES}\) >85%), were focal ( \(\:{F}_{PEZ\:}\) <23 mm), and close to the resection (<4mm) (Fig. 5 ). In contrast, for the poor outcome patient #37, the PEZs from all three classifiers showed low overlap with resection ( O RES 31mm). Finally, for poor outcome patient #44, although the PEZs were focal ( \(\:{F}_{PEZ}\) <20mm), they showed a low overlap with the resection ( O RES < \(\:41\%\) ) (Fig. 5 ). Discussion In this study, we introduce a novel supervised ML-based framework that delineates the EZ and predicts the surgical outcome in patients with DRE by extracting and combining features of interictal biomarkers from iEEG data in a fully automated fashion. Our framework detects spikes and ripples, extracts relevant temporal, spectral, and spatial features, and employs Random Forest classifiers to identify the PEZ. We evaluated the performance of both the individual features and Random Forest classifiers trained on SF, RF, and CF with RES as target. Our findings indicate that the SF and CF classifiers not only outperformed the models based on individual feature sets but also yielded a focal PEZ with high overlap and proximity to the resection in patients with good surgical outcome. Moreover, resection of the PEZ, defined by the CF, achieved an accuracy of ~ 80% in predicting surgical outcome, while resection of the clinically defined SOZ showed limited predictive value. This discrepancy likely reflects fundamental limitations of the SOZ concept in sEEG and depth-electrode-based evaluations, where seizure onset may be rapidly distributed across mesial and lateral temporal structures 66 , or reflect only the early propagation rather than the full epileptogenic network. 67 , 68 In addition, the SOZ is constrained by electrode coverage of the EZ, which may incompletely capture deeper or spatially remote regions contributing to seizure generation. As a result, SOZ localization alone may fail to identify all tissue necessary for resection and achievement of seizure freedom. Our framework generalizes across different implantation types and enables fully automated epileptogenic zone localization from interictal iEEG data, substantially reducing the need for manual identification and evaluation of epileptogenic brain regions. Spike spatial distribution plays a key role in defining the EZ Interictal biomarkers such as spikes and ripples, as well as their features including event rate, power, spike-ripple co-occurrence and their onsets, have been previously associated with the EZ. 11,58,69–72 Our results were consistent with these previous findings where several spike and ripple features had higher values within the resection (compared to outside) for good-outcome patients (Fig. 2 a). Notably, some features also showed elevated values within resection for patients with poor outcome but with lower effect size (Fig. 2 b). These findings indicate that although each feature independently captures distinct aspects of the EZ, resection of brain regions based on individual features alone does not necessarily lead to seizure freedom. Upon ranking feature importance, distance-based spike features emerged as the most informative (showing higher effect sizes) than other features in distinguishing epileptogenic from non-epileptogenic contacts (Fig. 2 b). This notion was further validated by the EZ prediction analysis, where performance metrics showed that the ED-SO, ED-HSR, and ED-HSP achieved higher AUC-ROC, AUC-PR, and \(\:{O}_{RES}\:\) values compared to the other features (Fig. 2 b). This aligns with a recent study showing that the spatial distribution of biomarkers plays a key role in defining the EZ. 48 Combined Features define the EZ better than individual features While each feature captures a distinct characteristic of the epileptogenic tissue, no single measure fully represents the complex spatial, spectral, and temporal dynamics of the EZ. A substantial body of literature suggests that ML-based integration of multiple electrophysiological biomarkers yields a more accurate localization of the EZ compared to individual features alone. 52 , 53 , 56 , 57 , 73 , 74 Such integration facilitates the discrimination between epileptogenic and non-epileptogenic contacts, even when measurements are incomplete or noisy. Our findings support this notion, as SF and CF classifiers outperformed individual features when predicting the epileptogenic and non-epileptogenic contacts. Specifically, SF and CF outperformed all individual features in terms of AUC-ROC and AUC-PR (except for the ED-HSR, ED-HSP, and ED-SO) (Fig. 3 a). Moreover, SF and CF exhibited higher \(\:{O}_{RES}\) than all individual features, except for the SRC (Fig. 3 b). Overall, our results highlight the added value of integrating multiple spatial, temporal, and spectral interictal features to improve the delineation of PEZ and support the notion that ML classifiers can capture nonlinear feature relationships of different epileptogenic activity more effectively than single-feature approaches. Ripple features provide modest improvement over spike features The Random Forest classifiers trained using the CF slightly outperformed those trained on SF, suggesting that spike features capture most of the relevant epileptogenic activity (Fig. 3 ). While combining spike and ripple features provided a more accurate representation of the PEZ than any individual feature alone, our results indicate that ripple features provided only a modest improvement over spike-based features. This may be due to the fact that not all ripples are pathological, underscoring the importance of distinguishing pathological from physiological ripples, which also occur in non-epileptogenic brain regions and during normal brain fucntion. 75 , 76 Although efforts have been made to address this distinction suggesting that pathological HFOs tend to co-occur with spikes and exhibit shorter durations and lower power and amplitude 22 , 75 , 77 , 78 , other studies reported the opposing findings. 79 , 80 Collectively, our findings suggest that, although integrating multiple biomarkers can offer incremental gains, spike features remain the dominant contributor to accurate identification of the PEZ based on interictal data. These findings are consistent with a recent study showing that low-frequency features alone were sufficient for the accurate delineation of the EZ and prediction of surgical outcome, with high-frequency features providing little or no additional value. 53 Combined Features predict outcome with higher accuracy than individual features Previous studies have reported strong associations between interictal biomarkers and postsurgical outcome. 11 , 22 , 37 , 81 – 86 These results indicate that, in patients with focal epilepsy, seizure-generating regions form a distinct pathological network isolated from surrounding brain tissue. When these pathological regions are completely resected, patients are more likely to achieve postoperative seizure freedom. Indeed, the spatial concordance between the identified epileptogenic regions and the resected zone has previously been used as a reliable predictor of success. 37 , 40 , 82 , 85 , 87 Our findings support this notion since we observed a decrease in the \(\:{O}_{RES}\) as well as increased \(\:{D}_{RES}\) in poor outcome patients (Fig. 4 ). These findings may indicate the presence of other epileptogenic regions distant from the resection, which may not have been resected due to overlap of the EZ with eloquent areas or could have been entirely missed [e.g., patient #37 and #44 (Fig. 5 )]. Upon assessing the ability of individual features to predict surgical outcome, we found that almost all features were predictive, except for the SPI, RR, and RPI (Fig. 2 b). All Random Forest classifiers were predictive of the surgical outcome (Fig. 4 b). When comparing outcome predictive performance across all individual features and classifiers, CF outperformed many of the features but not SR, ED-HSR, ED-HSP, ED-SO, ED-HRP, SF, and RF. These findings suggest that outcome prediction remains feasible using individual features alone, and that combining features does not necessarily improve performance, as many individual features are already informative (Fig. 4 c). Limitations Despite these promising results, several limitations should be acknowledged. Our retrospective study analyzed iEEG data from a single-center cohort of patients with DRE and heterogeneous pathologies. Because the models were trained using data from good outcome patients, future studies from larger, multicenter cohorts are needed to evaluate its clinical applicability, particularly within more homogeneous patient populations. Resection volumes from good outcome patients are often used as a gold standard for the EZ, despite the possibility that they include non-epileptogenic tissue. 1 , 88 This may result in either missing critical epileptogenic regions or defining areas larger than the actual EZ. Because ML model performance depends strongly on the accuracy of the ground truth used, improved approximations of the EZ are needed to provide more reliable targets for ML training. Moreover, iEEG is spatially limited since it is based on a subjective hypothesis agreed upon during presurgical evaluation. Consequently, the EZ was likely sampled well in good outcome patients but may have been missed in poor outcome patients. Advances in whole-brain iEEG implantations may overcome these drawbacks. Furthermore, future studies on longer data segments may be conducted to enhance the accuracy of PEZ in delineating the EZ. Finally, the retrospective nature of the study warrants further validation in larger, prospective cohorts to confirm the robustness and generalizability of these findings. Conclusion In this study, we systematically evaluated the predictive power of individual features and combined feature-based classifiers for localizing the EZ and predicting surgical outcome in patients with DRE. Our results demonstrate that Random Forest classifiers outperform individual features, providing more focal and accurate identification of the EZ. While spike features remain highly informative for outcome prediction, the integration of ripple features offers a modest additional advantage. Importantly, these findings support the use of short segments of interictal data to approximate the EZ and inform resection planning, potentially reducing reliance on ictal recordings. Future prospective studies with larger cohorts are necessary to validate and refine this approach for clinical translation. Materials and Methods Patient Cohort We retrospectively analyzed iEEG data of patients with DRE, from Boston Children’s Hospital (BCH, 51 patients, admitted between June 2011 and June 2023) and Cook Children’s Medical Center (CCMC, 11 patients, admitted between December 2019 and August 2021). The patients had long-term monitoring with iEEG and underwent resection neurosurgery. The selection of the patients was based on the following criteria: (i) availability of at least 5-minute interictal iEEG epochs with no biological or technical artifacts; (ii) availability of post-implantation computerized tomography (CT); (iii) availability of pre-operative and post-operative MRIs; (iv) accurate information about the resection volume, resected contacts, and the clinically defined SOZ; (v) availability of post-surgical outcome at least one year after surgery; (vi) data containing spikes and ripples; (vii) sampling frequency > 1,000 Hz. The surgical outcome was determined by a pediatric epileptologist after multiple follow-up visits. We used the Engel score 89 to classify patients as good (Engel 1) or poor outcome (Engel 2–4). The patient demographic and the clinical information of the cohort are provided in Table 1 . The protocol was approved by North Texas Regional IRB (2019 − 166; PI: C. Papadelis), that waived the need for informed consent considering the study’s retrospective nature. All methods and analyses were performed in accordance with relevant guidelines and regulations. Interictal iEEG recordings At BCH, long-term iEEG data were collected using stereotactic electrodes, subdural grids, or combined subdural grids and depth electrodes utilizing XLTEK Quantum NeuroWorks (Natus Inc., USA). The subdural electrodes had a diameter of 2–3 mm with a 10 mm inter-contact spacing, while the depth electrodes consisted of 6 to 16 contacts with a diameter of 0.8 mm and were linearly arranged approximately 3–5 mm apart. At CCMC, long-term iEEG data were collected using stereotactic electrodes utilizing Quantum NeuroWorks (Natus Inc., USA). The electrodes consisted of 8 to 16 contacts with a diameter of 0.8 mm and were linearly arranged approximately 3–5 mm apart. The data were recorded with a sampling frequency of 1,000 to 2,048 Hz. We specifically chose 5-minute interictal segments during non-REM slow-wave sleep (when applicable), occurring at least one hour before or after clinical seizures or half an hour before or after an electrographic seizure. 90 This approach ensured the incorporation of epochs with the highest spike rate and minimal motion artifacts. 91 , 92 Noisy channels and short segments with artifacts were excluded. Structural Imaging and iEEG recordings MRI scans utilizing standard magnetization-prepared rapid acquisition gradient-echo sequences were performed both before and following resection with a 3T scanner (TIM TRIO, Siemens AG at BCH and Magnetom Skyra at CCMC). Post-implantation CT scans, with voxel size set at 0.5 x 0.5 x 0.5 mm³, were performed after iEEG electrode implantation to document electrode locations. We aligned the electrode coordinates by registering the post-implantation CT scans with the preoperative MRI using Brainstorm . 93 We then projected the electrode locations onto the patient's cortical surface reconstructed from their preoperative MRI with Freesurfer . 94 We further adjusted the stereotactic and depth electrode positions to accommodate for brain shifting and pneumocephalus post-electrocorticography implantation. 95 The number and type of implanted electrodes for each patient are reported in Supplementary Tables 1 and 2 . Defining the SOZ and resection contacts For each patient, pediatric epileptologists determined the SOZ by visually examining ictal data. iEEG contacts that exhibited altered dynamics at the onset of seizures were designated as SOZ contacts. This process was conducted independently of the current study. To delineate the resection volume, we registered the preoperative and postoperative MRIs and manually outlined the boundaries of the resection volume on sequential slices using Brainstorm 93 and MATLAB’s volumeSegmentor toolbox. Subsequently, contacts ≤ 10 mm from resection were considered as resected (RES). Spike and Ripple Features To generate the features used in our classifiers, we focused on quantifying the spectral and spatiotemporal characteristics of spike and ripple propagations. We developed an in-house MATLAB algorithm to identify propagation sequences of spikes and ripples separately across all iEEG contacts for each patient. 65 After detecting the spikes, we employed our algorithm that marks the earliest spike within each sequence as onset (Fig. 1 b ) . 11 Then, it adds the next spike to the propagation sequence if it occurred within a predefined time window, i.e. the inter-event latency. The contacts in every propagation were indexed sequentially, starting from one, denoting the onset contact, followed by increasing numbers for the subsequent spikes. A similar approach was adopted for mapping ripple propagations (Fig. 1 c). Previous studies have shown that the inter-event latency for spikes and ripples is in the range of 3–15 ms 10,96 and 10–32 ms 10,42 , respectively. Based on these findings, we set here the inter-event latency to 10 and 30 ms, to map spike and ripple propagations respectively. 11 We also discarded propagations if more than 50% of their events occurred within 2 ms from each other. 52 , 97 Using spikes and their propagation events, we computed the following features for each channel (Fig. 1 d-f): (i) spike rate (SR); (ii) average spike power (SP); (iii) median spike propagation index (SPI); (iv) average Euclidean distance of contacts from the contact with the highest spike rate (ED-HSR); (v) average Euclidean distance of contacts from the contact with the highest average spike power (ED-HSP); and (vi) average Euclidean distance of the contacts from spike onset contact (ED-SO). Similarly, using ripples and their propagation events, we computed for each channel the: (i) ripple rate (RR); (ii) average ripple power (RP); (iii) median ripple propagation index (RPI); (iv) average Euclidean distance of contacts from the contact with the highest ripple rate (ED-HRR); (v) average Euclidean distance of contacts from the contact with the highest ripple power (ED-HRP); and (vi) average Euclidean distance of the contacts from ripple onset contact (ED-RO). Finally, we identified the cooccurrence of spikes and ripples in each contact and computed the average spike-ripple cooccurrence rate (SRC) per channel. Table 2 summarizes the features adopted in this study. We normalized the SR, RR, SP, RP and SRC between 0 and 1 using Min-Max scaling ensuring that the minimum value is scaled to 0 and the maximum to 1. For a feature \(\:x\) , the min-max normalized value \(\:x’\:\) is defined by: $$\:{x}^{{\prime\:}}=\frac{x-{x}_{min}}{{x}_{max}-{x}_{min}}$$ 1 where \(\:{x}_{min}\) is the minimum value and \(\:{x}_{max}\) is the maximum. The SPI, RPI, ED-HSR, ED-HRR, ED-HSP, ED-HRP, ED-SO, and ED-RO were scaled using Max-Min scaling ensuring that the maximum value is scaled to 0 and the minimum to 1. For a feature \(\:x\) , the max-min normalized value \(\:x’\:\) is defined by: $$\:x{\prime\:}=\frac{{x}_{max}-x}{{x}_{max}-{x}_{min}}$$ 2 Performance of individual features in EZ and outcome prediction For each feature, we estimated and compared the normalized values inside and outside the SOZ and resection for good and poor outcome patients, separately ( Wilcoxon signed-rank test). Then, to further evaluate the predictive power of individual features, we used the normalized values of each feature as predictors and RES as target. We performed the analysis only for good outcome patients since their resected tissue is assumed to contain the actual EZ. First, we assessed feature importance using one-way ANOVA. For each feature, we quantified the proportion of variance explained by the class labels (η²) 98 , which provides a measure of how strongly a feature differentiates between classes (i.e. epileptogenic vs. non-epileptogenic). Values of η² greater than 0.06 generally indicate that a feature has a meaningful influence on distinguishing between epileptogenic and non-epileptogenic classes. Next, for each feature, we used the LOPO-CV approach to assess its ability to predict the RES. At each step, one patient was left out, and the remaining patients’ feature values were collectively used to compute the optimal thresholds based on the maximum Youden index (J) obtained from the ROC curves. Then, the left-out patient’s features were thresholded using the computed optimal threshold to identify PEZ. We then defined the following: (i) true positives (TP), PEZ contacts that were located within the RES; (ii) false positives (FP), PEZ contacts that were located outside the RES; (iii) false negatives (FN), non-PEZ contacts that were located inside the RES; and (iv) true negatives (TN), non-PEZ contacts that were located outside the RES. We then calculated the following performance metrics: (i) sensitivity [TP/(TP + FN)]; (ii) specificity [TN/(TN + FP)]; (iii) precision [TP/(TP + FP)]; (iv) NPV [TN/(TN + FN)]; (v) accuracy [(TP + TN)/(TP + TN + FP + FN)]; (vi) F1-score [2 \(\:\times\:\) TP/(2 \(\:\times\:\) TP + FP + FN)]; (vii) AUC-ROC; (viii) AUC-PR. The overlap with resection ( \(\:{O}_{RES})\) was defined as the percentage of PEZ contacts within a 10 mm distance from the resection. The selection of the 10 mm cut-off was based on studies that showed that the gyral width is between 11–21 mm. 99 We hypothesized that the resection of PEZ is linked to favorable surgical outcome, thus, we considered \(\:{O}_{RES}\) as predictor and the dichotomized patient’s outcome as target. We considered: (i) TP, the number of good outcome patients who had ≥ 50% of the PEZ resected ( \(\:{O}_{RES}\:\) ≥ 50%); (ii) FN, the number of good outcome patients who had < 50% of the PEZ resected ( \(\:{O}_{RES}\:\) < 50%); (iii) FP, the number of poor outcome patients who had ≥ 50% of the PEZ resected; (iv) TN, the number of poor outcome patients who had < 50% of the PEZ resected. We then constructed the confusion matrices for each feature and computed the Fisher’s exact test p -values to find significance. Training and evaluation of Random Forest classifiers We trained and compared the performance of different Random Forest classifiers using different subsets of predictors and targets. We considered three sets of features as predictors: spike-features only (SF), ripple-features only (RF), and combined features (CF). We used RES to approximate the EZ. The setup resulted in training and evaluating three different classifiers with different combination of predictors (i.e., SF, RF, and CF). We exclusively trained and evaluated our classifiers using only patients with good outcome, since these patients remained seizure-free post-surgery, indicating successful removal of the EZ. Regarding the Random Forest classifier’s number of trees parameter, we conducted a five-fold cross-validation, varying the number of trees from 10 to 500 and selecting the value that yielded a stable AUC-ROC performance. Based on this analysis, we set the number of trees to 100 ( Supplementary Note 2 ). Classifier performance was evaluated using TP, TN, FP, and FN definitions identical to those used for individual feature analyses. Subsequently, we trained and tested the performance of the Random Forest classifier using LOPO-CV and computed performance metrics (i.e., sensitivity, specificity, precision, NPV, accuracy, F1-score, AUC-ROC, and AUC-PR). After estimating the PEZ using the trained classifiers, we computed F PEZ , O RES , and D RES . While training was performed using only good outcome patients, testing included both good and poor outcome patients. Robustness of PEZ across implantation types To evaluate whether our framework provides consistent findings among different implantation types, we examined the ability of PEZ to predict the EZ in good outcome patients with different implantation types. We first divided our good outcome patients into three subsets; (i) ECoG, patients with subdural implantations; (ii) sEEG, patients with stereotactic implantations; and (iii) ECoG and depth, patients with ECoG and a few depth electrodes ( Supplementary Table 1 ). For each subset, we trained and tested using CF with LOPO-CV and computed performance metrics (sensitivity, specificity, precision, NPV, accuracy, F1-score, AUC-ROC, AUC-PR) along with spatial properties ( \(\:{O}_{RES}\) , Dice score, \(\:{D}_{RES}\) , \(\:{F}_{PEZ}\) ) of the PEZ. Random forest classifiers performance to predict surgical outcome We presumed that the \(\:{O}_{RES}\) of PEZ predicted from the classifiers SF, RF, and CF could predict outcome. Specifically, we assumed that in patients with good outcome, the PEZ would be located within resection, whereas in those with poor outcome, the resection would have missed the EZ. Subsequently, we performed a prediction analysis where we used the \(\:{O}_{RES}\) of the PEZ as predictor of outcome. We considered good and poor outcomes as positive and negative classes, respectively. We then defined: (i) TP, the number of good outcome patients who had ≥ 50% of the PEZ resected ( \(\:{O}_{RES}\) ≥ 50%) ; (ii) FN, the number of good outcome patients who had < 50% of the PEZ resected ( \(\:{O}_{RES}\) < 50%); (iii) FP, the number of poor outcome patients who had ≥ 50% of the PEZ resected; and (iv) TN, the number of poor outcome patients who had < 50% of the PEZ resected. We constructed confusion matrices for each feature set and computed the Fisher’s exact test P -values to assess significance. Finally, we computed the sensitivity, specificity, precision, NPV, F1-score, and accuracy of the outcome predictions. Statistical Analysis The Kolmogorov-Smirnov test was used to test the normality of features. Cliff’s d measure was used to compute effect sizes. Demographic data and PEZ spatial measures ( \(\:{F}_{PEZ}\) , \(\:{O}_{RES}\) , Dice score, and \(\:{D}_{RES}\) ) are reported as median (25th -75th percentile). The chi-squared test was used to compare the effect of different cohort characteristics (i.e., gender, implantation side, epilepsy localization, and MRI findings) on outcome. We applied two-sided non-parametric Wilcoxon signed-rank test for all paired comparisons (median feature values inside vs. outside the SOZ and resection). We applied the two-sided Wilcoxon rank-sum test for non-paired comparisons between good and poor outcome patients. Bonferroni correction was applied in all multiple comparison tests. We used the one-sided Fisher’s exact test to evaluate the predictive value of the PEZs to predict outcome. To evaluate the performance of all individual features and trained Random Forest classifier models, we performed a comparative analysis of EZ prediction using the AUC-ROC, AUC-PR, and \(\:{O}_{RES}\) . Pairwise comparisons were performed using the Wilcoxon signed-rank test on each performance metric, trained and evaluated using the LOPO-CV approach. False discovery Rate (FDR) was applied to correct for multiple comparisons. The results are displayed in a square matrix where each element indicates whether a significant difference ( P < 0.05) between the different predictor models was observed and, if so, identifies the model with the higher median value. For outcome prediction using individual features, classifier models, and SOZ, we performed a pairwise comparison using the McNemar test. The elements in the corresponding square matrix show both the presence of significant differences ( P < 0.05) and the ID of the model with higher accuracy. The McNemar test was used to examine differences in the ability to predict outcomes using confusion matrices between different features and the proposed ML classifiers. All analysis was performed with MATLAB 2023b (The MathWorks, Inc.). Declarations Funding This research was supported by the National Institute of Neurological Disorders & Stroke (R01NS104116-01A1 and R01NS134944, PI: C. Papadelis). Data availability The data are available from the corresponding author upon request. Code availability The code is available from the corresponding author upon request. Author Contributions H.P., S.J., and C.P. conceptualized the project and experimental design. H.P. and S.J. developed the mathematical formulation and the codes of the proposed framework. H.P., S.J., S.P., E.T., J.M., J.B., S.S., P.P., and C.P., contributed to the acquisition and analysis of data. H.P. and S.J. participated in the development of various mathematical formulations, statistical analysis, and representation of results of the study. H.P., S.J., and C.P. contributed to drafting the text and preparing figures. All authors have read and approved the manuscript. Competing Interests The authors have no competing interests to disclose. References Jehi, L. The Epileptogenic Zone: Concept and Definition. Epilepsy Curr. 18 , 12–16 (2018). Rosenow, F. & Lüders, H. Presurgical evaluation of epilepsy. Brain 124 , 1683–1700 (2001). Lai, N., Li, Z., Xu, C., Wang, Y. & Chen, Z. Diverse nature of interictal oscillations: EEG-based biomarkers in epilepsy. Neurobiol. Dis. 177 , 105999 (2023). Emmady, P. D., Asuncion, R. M. D. & Anilkumar, A. C. EEG Abnormal Waveforms. in StatPearls (StatPearls Publishing, 2025). Widjaja, E. et al. Cost-effectiveness of pediatric epilepsy surgery compared to medical treatment in children with intractable epilepsy. Epilepsy Res. 94 , 61–68 (2011). Papadelis, C. & Perry, M. S. Localizing the Epileptogenic Zone with Novel Biomarkers. Semin Pediatr. Neurol. 39 , 100919 (2021). Engel, J. J. et al. Epilepsy biomarkers. Epilepsia 54 (Suppl 4), 61–69 (2013). Staley, K. J., White, A. & Dudek, F. E. Interictal spikes: harbingers or causes of epilepsy? Neurosci. Lett. 497 , 247–250 (2011). Staley, K. J. & Dudek, F. E. Interictal spikes and epileptogenesis. Epilepsy Curr. 6 , 199–202 (2006). Tamilia, E. et al. Surgical resection of ripple onset predicts outcome in pediatric epilepsy. Ann. Neurol. 84 , 331–346 (2018). Jahromi, S. et al. Overlap of spike and ripple propagation onset predicts surgical outcome in epilepsy. Ann. Clin. Transl Neurol. 11 , 2530–2547 (2024). Klotz, K. A., Sag, Y., Schönberger, J. & Jacobs, J. Scalp Ripples Can Predict Development of Epilepsy After First Unprovoked Seizure in Childhood. Ann. Neurol. 89 , 134–142 (2021). Staba, R. J., Wilson, C. L., Bragin, A., Fried, I. & Engel, J. J. Quantitative analysis of high-frequency oscillations (80–500 Hz) recorded in human epileptic hippocampus and entorhinal cortex. J. Neurophysiol. 88 , 1743–1752 (2002). Jacobs, J. et al. Interictal high-frequency oscillations (80–500 Hz) are an indicator of seizure onset areas independent of spikes in the human epileptic brain. Epilepsia 49 , 1893–1907 (2008). Höller, Y. et al. High-frequency oscillations in epilepsy and surgical outcome. A meta-analysis. Front. Hum. Neurosci. 9 , 574 (2015). González Otárula, K. A., von Ellenrieder, N., Cuello-Oderiz, C., Dubeau, F. & Gotman, J. High-Frequency Oscillation Networks and Surgical Outcome in Adult Focal Epilepsy. Ann. Neurol. 85 , 485–494 (2019). Bragin, A., Engel, J. J., Wilson, C. L., Fried, I. & Mathern, G. W. Hippocampal and entorhinal cortex high-frequency oscillations (100–500 Hz) in human epileptic brain and in kainic acid–treated rats with chronic seizures. Epilepsia 40 , 127–137 (1999). Weiss, S. A. et al. Localizing epileptogenic regions using high-frequency oscillations and machine learning. Biomark. Med. 13 , 409–418 (2019). Roehri, N. et al. High-frequency oscillations are not better biomarkers of epileptogenic tissues than spikes. Ann. Neurol. 83 , 84–97 (2018). Wang, S. et al. Ripple classification helps to localize the seizure-onset zone in neocortical epilepsy. Epilepsia 54 , 370–376 (2013). Matsumoto, A. et al. Pathological and physiological high-frequency oscillations in focal human epilepsy. J. Neurophysiol. 110 , 1958–1964 (2013). Fabbri, L. et al. Spikes on ripples are better interictal biomarkers of epilepsy than spikes or ripples. Brain Commun. 7 , fcaf056 (2025). Zijlmans, M. et al. Ictal and interictal high frequency oscillations in patients with focal epilepsy. Clin. Neurophysiol. 122 , 664–671 (2011). Salami, P. et al. Dynamics of interictal spikes and high-frequency oscillations during epileptogenesis in temporal lobe epilepsy. Neurobiol. Dis. 67 , 97–106 (2014). Shi, W. et al. Spike ripples localize the epileptogenic zone best: an international intracranial study. Brain 147 , 2496–2506 (2024). Frauscher, B. et al. High-frequency oscillations: The state of clinical research. Epilepsia 58 , 1316–1329 (2017). Jacobs, J. et al. High-frequency electroencephalographic oscillations correlate with outcome of epilepsy surgery. Ann. Neurol. 67 , 209–220 (2010). Frauscher, B. et al. High-Frequency Oscillations in the Normal Human Brain. Ann. Neurol. 84 , 374–385 (2018). Engel, J. J., Bragin, A., Staba, R. & Mody, I. High-frequency oscillations: what is normal and what is not? Epilepsia 50, 598–604 (2009). Zweiphenning, W. J. E. M. et al. Correcting for physiological ripples improves epileptic focus identification and outcome prediction. Epilepsia 63 , 483–496 (2022). Bruder, J. C. et al. Physiological Ripples Associated with Sleep Spindles Differ in Waveform Morphology from Epileptic Ripples. Int. J. Neural Syst. 27 , 1750011 (2017). Shi, W. et al. Spike ripples localize the epileptogenic zone best: an international intracranial study. Brain 147 , 2496–2506 (2024). Amiri, M., Lina, J. M., Pizzo, F. & Gotman, J. High Frequency Oscillations and spikes: Separating real HFOs from false oscillations. Clin. Neurophysiol. 127 , 187–196 (2016). Jacobs, J. et al. The identification of distinct high-frequency oscillations during spikes delineates the seizure onset zone better than high-frequency spectral power changes. Clin. Neurophysiol. 127 , 129–142 (2016). Alamoudi, O. A., Ilyas, A., Pati, S. & Iasemidis, L. Interictal localization of the epileptogenic zone: Utilizing the observed resonance behavior in the spectral band of surrounding inhibition. Frontiers Neuroscience 16–2022 , (2022). Heers, M. et al. Spectral bandwidth of interictal fast epileptic activity characterizes the seizure onset zone. NeuroImage: Clin. 17 , 865–872 (2018). Partamian, H. et al. Machine learning on interictal intracranial EEG predicts surgical outcome in drug resistant epilepsy. npj Digit. Med. 8 , 138 (2025). Worrell, G. A. et al. Recording and analysis techniques for high-frequency oscillations. Prog Neurobiol. 98 , 265–278 (2012). Mercier, M. R. et al. Advances in human intracranial electroencephalography research, guidelines and good practices. NeuroImage 260 , 119438 (2022). Matarrese, M. A. G. et al. Spike propagation mapping reveals effective connectivity and predicts surgical outcome in epilepsy. Brain 146 , 3898–3912 (2023). Diamond, J. M., Chapeton, J. I., Theodore, W. H., Inati, S. K. & Zaghloul, K. A. The seizure onset zone drives state-dependent epileptiform activity in susceptible brain regions. Clin. Neurophysiol. 130 , 1628–1641 (2019). Tamilia, E. et al. Noninvasive Mapping of Ripple Onset Predicts Outcome in Epilepsy Surgery. Ann. Neurol. 89 , 911–925 (2021). Alarcon, G. et al. Origin and propagation of interictal discharges in the acute electrocorticogram. Implications for pathophysiology and surgical treatment of temporal lobe epilepsy. Brain 120 , 2259–2282 (1997). Lin, J. et al. High frequency oscillation network dynamics predict outcome in non-palliative epilepsy surgery. Brain Commun. 6 , fcae032 (2024). Conrad, E. C. et al. Spatial distribution of interictal spikes fluctuates over time and localizes seizure onset. Brain 143 , 554–569 (2019). Gliske, S. V. et al. Variability in the location of high frequency oscillations during prolonged intracranial EEG recordings. Nat. Commun. 9 , 2155 (2018). Cimbalnik, J. et al. Multi-feature localization of epileptic foci from interictal, intracranial EEG. Clin. Neurophysiol. 130 , 1945–1953 (2019). Jaber, K. et al. A spatial perturbation framework to validate implantation of the epileptogenic zone. Nat. Commun. 15 , 5253 (2024). Jiang, X., Cai, Z., Gonsisko, C., Worrell, G. A. & He, B. Mapping epileptogenic brain using a unified spatial–temporal–spectral source imaging framework. Proceedings of the National Academy of Sciences 122, e2510015122 (2025). Mirza, B. et al. Machine Learning and Integrative Analysis of Biomedical Big Data. Genes (Basel) 10 , (2019). Zitnik, M. et al. Machine Learning for Integrating Data in Biology and Medicine: Principles, Practice, and Opportunities. Inf. Fusion . 50 , 71–91 (2019). Shamas, M., Yeh, H. J., Fried, I., Engel, J. J. & Staba, R. J. High-rate leading spikes in propagating spike sequences predict seizure outcome in surgical patients with temporal lobe epilepsy. Brain Commun. 5 , fcad289 (2023). Klimes, P. et al. Interictal stereo-electroencephalography features below 45 Hz are sufficient for correct localization of the epileptogenic zone and postsurgical outcome prediction. Epilepsia 65 , 2935–2945 (2024). Raghu, A. et al. Machine Learning-Based localization of the epileptogenic zone using High-Frequency oscillations from SEEG: A Real-World approach. J. Clin. Neurosci. 135 , 111177 (2025). Daida, A. et al. AI-based localization of the epileptogenic zone using intracranial EEG. Epilepsia Open. https://doi.org/10.1002/epi4.70130 (2025). Wang, Y. et al. Automatic Localization of Seizure Onset Zone Based on Multi-Epileptogenic Biomarkers Analysis of Single-Contact from Interictal SEEG. Bioengineering 9, (2022). Varatharajah, Y. et al. Integrating artificial intelligence with real-time intracranial EEG monitoring to automate interictal identification of seizure onset zones in focal epilepsy. J. Neural Eng. 15 , 046035 (2018). Varatharajah, Y. et al. Inter-ictal Seizure Onset Zone localization using unsupervised clustering and Bayesian Filtering. in. 8th International IEEE/EMBS Conference on Neural Engineering (NER) 533–539 (2017). (2017). 10.1109/NER.2017.8008407 Islam, M. R., Zhao, X., Miao, Y., Sugano, H. & Tanaka, T. Epileptic seizure focus detection from interictal electroencephalogram: a survey. Cogn. Neurodyn. 17 , 1–23 (2023). Cimbalnik, J. et al. Multi-feature localization of epileptic foci from interictal, intracranial EEG. Clin. Neurophysiol. 130 , 1945–1953 (2019). Schönberger, J. et al. Distinction of Physiologic and Epileptic Ripples: An Electrical Stimulation Study. Brain Sci 11 , (2021). Janca, R. et al. Detection of interictal epileptiform discharges using signal envelope distribution modelling: application to epileptic and non-epileptic intracranial recordings. Brain Topogr . 28 , 172–183 (2015). Weiss, S. A. et al. Fast ripples reflect increased excitability that primes epileptiform spikes. Brain Commun. 5 , fcad242 (2023). Cho, J. R. et al. Resection of individually identified high-rate high-frequency oscillations region is associated with favorable outcome in neocortical epilepsy. Epilepsia 55 , 1872–1883 (2014). Jahromi, S. et al. Overlap of spike and ripple propagation onset predicts surgical outcome in epilepsy. Ann. Clin. Transl Neurol. 11 , 2530–2547 (2024). Andrews, J. P. et al. Association of Seizure Spread With Surgical Failure in Epilepsy. JAMA Neurol. 76 , 462–469 (2019). Andrews, J. P. et al. Early seizure spread and epilepsy surgery: A systematic review. Epilepsia 61 , 2163–2172 (2020). Li, A. et al. Neural fragility as an EEG marker of the seizure onset zone. Nat. Neurosci. 24 , 1465–1474 (2021). Corona, L. et al. Non-invasive mapping of epileptogenic networks predicts surgical outcome. Brain 146 , 1916–1931 (2023). Matarrese, M. A. G. et al. Spike propagation mapping reveals effective connectivity and predicts surgical outcome in epilepsy. Brain 146 , 3898–3912 (2023). Gunnarsdottir, K. M. et al. Source-sink connectivity: a novel interictal EEG marker for seizure localization. Brain 145 , 3901–3915 (2022). Islam, M. R., Zhao, X., Miao, Y., Sugano, H. & Tanaka, T. Epileptic seizure focus detection from interictal electroencephalogram: a survey. Cogn. Neurodyn. 17 , 1–23 (2023). Miao, Y., Iimura, Y., Sugano, H., Fukumori, K. & Tanaka, T. Seizure onset zone identification using phase-amplitude coupling and multiple machine learning approaches for interictal electrocorticogram. Cogn. Neurodyn. 17 , 1591–1607 (2023). Akter, M. S. et al. Statistical Features in High-Frequency Bands of Interictal iEEG Work Efficiently in Identifying the Seizure Onset Zone in Patients with Focal Epilepsy. Entropy 22, (2020). Fabbri, L. et al. Noninvasive classification of physiological and pathological high frequency oscillations in children. Brain Commun. 7 , fcaf170 (2025). Mooij, A. H. et al. Accurate differentiation between physiological and pathological ripples recorded with scalp-EEG. Clin. Neurophysiol. 143 , 172–181 (2022). Karpychev, V. et al. Epileptogenic high-frequency oscillations present larger amplitude both in mesial temporal and neocortical regions. Front. Hum. Neurosci. 16 , 984306 (2022). Cimbalnik, J. et al. Physiological and pathological high frequency oscillations in focal epilepsy. Ann. Clin. Transl. Neurol. 5 , 1062–1076 (2018). Alkawadri, R. et al. The spatial and signal characteristics of physiologic high frequency oscillations. Epilepsia 55 , 1986–1995 (2014). Chen, C. et al. A region-specific modulation of sleep slow waves on interictal epilepsy markers in focal epilepsy. Epilepsia 64 , 973–985 (2023). Lundstrom, B. N., Brinkmann, B. H. & Worrell, G. A. Low frequency novel interictal EEG biomarker for localizing seizures and predicting outcomes. Brain Commun. 3 , fcab231 (2021). Corona, L. et al. Non-invasive mapping of epileptogenic networks predicts surgical outcome. Brain 146 , 1916–1931 (2023). Shah, P. et al. High interictal connectivity within the resection zone is associated with favorable post-surgical outcomes in focal epilepsy patients. Neuroimage Clin. 23 , 101908 (2019). Gunnarsdottir, K. M. et al. Source-sink connectivity: a novel interictal EEG marker for seizure localization. Brain 145 , 3901–3915 (2022). Rijal, S. et al. Functional connectivity discriminates epileptogenic states and predicts surgical outcome in children with drug resistant epilepsy. Sci. Rep. 13 , 9622 (2023). Jacobs, J. et al. Removing high-frequency oscillations: A prospective multicenter study on seizure outcome. Neurology 91 , e1040–e1052 (2018). Tamilia, E. et al. Surgical resection of ripple onset predicts outcome in pediatric epilepsy. Ann. Neurol. 84 , 331–346 (2018). Nissen, I. A. et al. Optimization of epilepsy surgery through virtual resections on individual structural brain networks. Sci. Rep. 11 , 19025 (2021). Engel, J. Jr Surgery for seizures. N. Engl. J. Med. 334 , 647–653 (1996). Dimakopoulos, V. et al. Protocol for multicentre comparison of interictal high-frequency oscillations as a predictor of seizure freedom. Brain Commun. 4 , fcac151 (2022). Frauscher, B. et al. Facilitation of epileptic activity during sleep is mediated by high amplitude slow waves. Brain 138 , 1629–1641 (2015). Sammaritano, M., Gigli, G. L. & Gotman, J. Interictal spiking during wakefulness and sleep and the localization of foci in temporal lobe epilepsy. Neurology 41 , 290–290 (1991). Tadel, F., Baillet, S., Mosher, J. C., Pantazis, D. & Leahy, R. M. Brainstorm: a user-friendly application for MEG/EEG analysis. Computational intelligence and neuroscience 1–13 (2011). (2011). de Rodrigues, M. et al. K. A FreeSurfer-Compliant Consistent Manual Segmentation of Infant Brains Spanning the 0–2 Year Age Range. Front. Hum. Neurosci. 9, 21. (2015). Roberts, D. W., Hartov, A., Kennedy, F. E., Miga, M. I. & Paulsen, K. D. Intraoperative brain shift and deformation: a quantitative analysis of cortical displacement in 28 cases. Neurosurgery 43 , 749–758 (1998). Azeem, A. et al. Interictal spike networks predict surgical outcome in patients with drug-resistant focal epilepsy. Ann. Clin. Transl Neurol. 8 , 1212–1223 (2021). Conrad, E. C. et al. Spike patterns surrounding sleep and seizures localize the seizure-onset zone in focal epilepsy. Epilepsia 64 , 754–768 (2023). Maher, J. M., Markey, J. C. & Ebert-May, D. The Other Half of the Story: Effect Size Analysis in Quantitative Research. CBE—Life Sci. Educ. 12 , 345–351 (2013). Ono, M., Kubik, S. & Abernathey, C. D. Atlas of the Cerebral Sulci (G. Thieme, 1990). Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterial.docx Cite Share Download PDF Status: Published Journal Publication published 26 Mar, 2026 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Revision requested 25 Feb, 2026 Reviews received at journal 22 Feb, 2026 Reviews received at journal 21 Feb, 2026 Reviewers agreed at journal 13 Feb, 2026 Reviewers agreed at journal 12 Feb, 2026 Reviewers agreed at journal 12 Feb, 2026 Reviewers invited by journal 11 Feb, 2026 Editor assigned by journal 11 Feb, 2026 Editor invited by journal 03 Feb, 2026 Submission checks completed at journal 29 Jan, 2026 First submitted to journal 29 Jan, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8682213","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":591207590,"identity":"2387dcdf-bb08-47e2-9563-e85574d6eaf2","order_by":0,"name":"Hmayag Partamian","email":"","orcid":"","institution":"Cook Children's Health Care System","correspondingAuthor":false,"prefix":"","firstName":"Hmayag","middleName":"","lastName":"Partamian","suffix":""},{"id":591207593,"identity":"df80f4e1-e2d4-41d9-9574-8ca4d5a473b4","order_by":1,"name":"Saeed Jahromi","email":"","orcid":"","institution":"Cook Children's Health Care System","correspondingAuthor":false,"prefix":"","firstName":"Saeed","middleName":"","lastName":"Jahromi","suffix":""},{"id":591207594,"identity":"920ae43c-eeb1-4e5c-889f-962ae2247738","order_by":2,"name":"M Scott Perry","email":"","orcid":"","institution":"Cook Children's Health Care System","correspondingAuthor":false,"prefix":"","firstName":"M","middleName":"Scott","lastName":"Perry","suffix":""},{"id":591207595,"identity":"af9d8026-a3f0-49af-8d35-39a74f93af03","order_by":3,"name":"Eleonora Tamilia","email":"","orcid":"","institution":"Boston Children's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Eleonora","middleName":"","lastName":"Tamilia","suffix":""},{"id":591207596,"identity":"0f5c9299-5f65-4db6-8607-6f9fb3f5e882","order_by":4,"name":"Joseph R. Madsen","email":"","orcid":"","institution":"Boston Children's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Joseph","middleName":"R.","lastName":"Madsen","suffix":""},{"id":591207598,"identity":"fc005e75-835a-4c38-aad2-3b1916c14641","order_by":5,"name":"Jeffrey Bolton","email":"","orcid":"","institution":"Boston Children's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Jeffrey","middleName":"","lastName":"Bolton","suffix":""},{"id":591207599,"identity":"a2977f44-1be7-4627-926b-9e4c9b73efd7","order_by":6,"name":"Scellig Stone","email":"","orcid":"","institution":"Boston Children's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Scellig","middleName":"","lastName":"Stone","suffix":""},{"id":591207600,"identity":"6ddbd252-4b3f-46a5-8c4e-df790cba6e54","order_by":7,"name":"Phillip Pearl","email":"","orcid":"","institution":"Boston Children's Hospital","correspondingAuthor":false,"prefix":"","firstName":"Phillip","middleName":"","lastName":"Pearl","suffix":""},{"id":591207601,"identity":"cdf917ab-22d4-45b5-8231-dd8154bcfd9b","order_by":8,"name":"Christos Papadelis","email":"data:image/png;base64,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","orcid":"","institution":"Cook Children's Health Care System","correspondingAuthor":true,"prefix":"","firstName":"Christos","middleName":"","lastName":"Papadelis","suffix":""}],"badges":[],"createdAt":"2026-01-23 19:53:50","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8682213/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8682213/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-026-45177-x","type":"published","date":"2026-03-26T16:12:21+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":102852278,"identity":"6f815127-c55e-4e29-a0f1-7a0b60e50495","added_by":"auto","created_at":"2026-02-17 14:36:06","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":741647,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eProcessing pipeline of the automatic epileptogenic zone (EZ) classifier. a\u003c/strong\u003e Five-minute interictal iEEG segments were preprocessed, cleaned, and filtered in two frequency bands: spike band (1–70 Hz) and ripple band (80–250 Hz). Automatic spike and ripple detectors identified spike and ripple events, from which temporal, spatial, and spectral features were extracted. The seizure onset zone (SOZ) and resection zone (RES) were defined based on clinical reports and postoperative MRI-defined resection volumes. Multiple Random Forest classifiers were trained using different feature and label sets for mapping the predicted EZ (PEZ), and the correspondence between the PEZ and surgical outcome was evaluated. \u003cstrong\u003eb\u003c/strong\u003e In the spike band, the automatic spike detector identified spikes (blue). Detected propagation events (\u003cem\u003ep1–p3\u003c/em\u003e) were analyzed to compute a spike propagation index per channel. Channels exhibiting earliest activation within each event were defined as onset spikes (red signal in \u003cem\u003ep2\u003c/em\u003e). \u003cstrong\u003ec\u003c/strong\u003e In the ripple band, the automatic ripple detector identified ripples (yellow). Similar propagation analysis of detected events (\u003cem\u003ep4–p7\u003c/em\u003e) yielded a ripple propagation index per channel, and channels showing earliest activation were defined as onset ripples (red signal in p7). \u003cstrong\u003ed\u003c/strong\u003e Spectral power features were computed for spikes and ripples across all channels and normalized within [0,1].\u003cstrong\u003e e\u003c/strong\u003e Spike and ripple rates were calculated per channel to characterize interictal activity. \u003cstrong\u003ef\u003c/strong\u003e Spatial distribution features were derived using Euclidean distances between each channel involved in the events and the contacts with highest event rate, maximum power, and onset activity. \u003cstrong\u003eg\u003c/strong\u003e Coregistrationof pre- and postoperative MRI with CT scans containing intracranial implantations was used to localize iEEG electrode coordinates and delineate the resected volume from postoperative MRI, which served as the gold standard for epileptogenic zone (EZ) prediction, with the contacts located within the resection defined as resected contacts (RES). \u003cstrong\u003eh \u003c/strong\u003eMultiple Random Forest classifiers were trained using three different feature sets [spike features (SF), ripple features (RF), and combined features (CF)] to classify contacts as epileptogenic (EP) or non-epileptogenic (non-EP).\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-8682213/v1/eb9672d03094b2ac0484bf20.png"},{"id":102963068,"identity":"fd9f517c-c00b-4bd3-b299-f077b51d5cc6","added_by":"auto","created_at":"2026-02-19 04:13:09","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":337635,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePredictive power of individual features. a \u003c/strong\u003eComparison of feature values inside versus outside the resection zone in good and poor outcome patients. For each feature, the red box plot represents the median feature value inside the resection, and the blue box plot represents the median value outside the resection. The notches indicate the median across all patients, while the lower and upper box edges correspond to the 25\u003csup\u003eth\u003c/sup\u003e and 75\u003csup\u003eth\u003c/sup\u003e percentiles. Whiskers extend to the minimum and maximum values after omitting outliers. Significant differences are indicated by asterisks: *\u003cem\u003eP\u003c/em\u003e\u0026nbsp;\u0026lt; 0.05, **\u003cem\u003eP\u003c/em\u003e\u0026nbsp;\u0026lt; 0.01, ***\u003cem\u003eP\u003c/em\u003e\u0026nbsp;\u0026lt; 0.001, ****\u003cem\u003eP\u003c/em\u003e\u0026nbsp;\u0026lt; 0.0001 (\u003cem\u003eWilcoxon signed-rank\u003c/em\u003e test, significance levels were corrected for multiple comparisons using the Bonferroni method). The number displayed below the asterisks represents the effect size. \u003cstrong\u003eb\u003c/strong\u003e Ranked importance (effect size) and performance of individual features in predicting the resection (RES) in good outcome patients.\u0026nbsp; Performance measures were evaluated using Youden index–based optimal thresholding with leave-one-patient-out cross-validation (LOPO-CV). Asterisks denote statistical significance of features in predicting surgical outcome based on the \u003cem\u003eFisher\u003c/em\u003e test (*\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05). Features: SR (spike rate), SP (spike power), SPI (spike propagation index), ED-HSP (distance from highest spike power), ED-HSR (distance from highest spike rate), ED-SO (distance from spike onset), RR (ripple rate), RP (ripple power), RPI (ripple propagation index), ED-HRP (distance from highest ripple power), ED-HRR (distance from highest ripple rate), ED-RO (distance from ripple onset), SRC (spike-ripple co-occurrence).\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-8682213/v1/08a32e1f43dc23d9cabe0bfd.png"},{"id":102852281,"identity":"48455a2e-3108-4a12-9f34-4991bf04e5ed","added_by":"auto","created_at":"2026-02-17 14:36:06","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":664362,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePredicting the epileptogenic zone (EZ) using individual features and different feature sets. a \u003c/strong\u003ePerformance metrics were evaluated for different classifiers in predicting the EZ (PEZ) using individual features and spike features (SF), ripple features (RF), and combined features (CF). Metrics included area under the receiver operating characteristic curve (AUC-ROC), area under the precision–recall curve (AUC-PR), and the overlap with resection (). Classification performance was assessed using a leave-one-patient-out cross-validation (LOPO-CV) approach trained on good outcome patients. The notches represent the median metric value across all patients, while the lower and upper box edges correspond to the 25\u003csup\u003eth\u003c/sup\u003e and 75\u003csup\u003eth\u003c/sup\u003e percentiles. Whiskers extend to the minimum and maximum values after excluding outliers. The number above each boxplot represents the median value.\u003cstrong\u003e b\u003c/strong\u003e Pairwise comparison of AUC-ROC, AUC-PR, and \u0026nbsp;for epileptogenic zone prediction, derived from individual interictal features and machine-learning classifiers using SF, RF, and CF as predictors.\u003cstrong\u003e \u003c/strong\u003eEach red cell indicates a significant difference (\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05, \u003cem\u003eWilcoxon signed-rank\u003c/em\u003e test, FDR corrected) between the corresponding pair of models, with the ID of the better-performing model (higher median AUC-ROC, AUC-PR, and , respectively) displayed inside the cell.\u003cstrong\u003e \u003c/strong\u003eBlue cells indicate no differences. Features: SR (spike rate), SP (spike power), SPI (spike propagation index), ED-HSP (distance from highest spike power), ED-HSR (distance from highest spike rate), ED-SO (distance from spike onset), RR (ripple rate), RP (ripple power), RPI (ripple propagation index), ED-HRP (distance from highest ripple power), ED-HRR (distance from highest ripple rate), ED-RO (distance from ripple onset), SRC (spike-ripple co-occurrence).\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-8682213/v1/631c5e121dc226bc10d4b022.png"},{"id":102963777,"identity":"c2b4be9c-b844-4225-af90-d0ac7a8a18ab","added_by":"auto","created_at":"2026-02-19 04:20:32","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":291049,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSpatial properties of the predicted epileptogenic zone (PEZ) and outcome prediction. a\u003c/strong\u003e Overlap with resection (, %), distance from resection (, mm), and focality (, mm) were computed for PEZ identified by classifiers trained using spike features (SF), ripple features (RF), and combined features (CF) as predictors for good- and poor outcome patients. The notches indicate the median values across all patients, while the lower and upper box edges correspond to the 25\u003csup\u003eth\u003c/sup\u003e and 75\u003csup\u003eth\u003c/sup\u003e percentiles. Whiskers extend to the minimum and maximum values after omitting outliers. Significant differences are indicated by asterisks: *\u003cem\u003eP\u003c/em\u003e\u0026nbsp;\u0026lt; 0.05, **\u003cem\u003eP\u003c/em\u003e\u0026nbsp;\u0026lt; 0.01, ***\u003cem\u003eP\u003c/em\u003e\u0026nbsp;\u0026lt; 0.001, ****\u003cem\u003eP\u003c/em\u003e\u0026nbsp;\u0026lt; 0.0001 (\u003cem\u003eWilcoxon rank-sum \u003c/em\u003etest, significance levels were corrected for multiple comparisons using the Bonferroni method). \u003cstrong\u003eb\u003c/strong\u003e Outcome prediction based on resection of the PEZ.\u003cstrong\u003e \u003c/strong\u003eConfusion matrices show outcome prediction when 50% of the PEZ was resected. The PEZ was predicted using spike features (SF), ripple features (RF), and combined features (CF). The seizure onset zone (SOZ) was also assessed for outcome prediction. Evaluation metrics included sensitivity, specificity, precision, negative predictive value (NPV), F1-score, and accuracy. Significant differences are indicated by asterisk above the confusion matrices: *\u003cem\u003eP\u003c/em\u003e\u0026nbsp;\u0026lt; 0.05, **\u003cem\u003eP\u003c/em\u003e\u0026nbsp;\u0026lt; 0.01, ***\u003cem\u003eP\u003c/em\u003e\u0026nbsp;\u0026lt; 0.001, ****\u003cem\u003eP\u003c/em\u003e\u0026nbsp;\u0026lt; 0.0001 (\u003cem\u003eFisher’s exact\u003c/em\u003e\u0026nbsp;test). \u003cstrong\u003ec\u003c/strong\u003e Pairwise comparison of individual feature and feature subset-based classifiers (SF, RF, and CF) outcome prediction using the McNemar test.\u003cstrong\u003e \u003c/strong\u003eEach yellow cell indicates a significant difference (p \u0026lt; 0.05, \u003cem\u003eMcNemar\u003c/em\u003e test) between the corresponding pair of models, with the ID of the better-performing model (higher average prediction) displayed inside the cell.\u003cstrong\u003e \u003c/strong\u003eBlue cells indicate no differences. Feature sets: SF (spike features), RF (ripple features), CF (combined features). Features: SR (spike rate), SP (spike power), SPI (spike propagation index), ED-HSP (distance from highest spike power), ED-HSR (distance from highest spike rate), ED-SO (distance from spike onset), RR (ripple rate), RP (ripple power), RPI (ripple propagation index), ED-HRP (distance from highest ripple power), ED-HRR (distance from highest ripple rate), ED-RO (distance from ripple onset), and SRC (spike–ripple co-occurrence).\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-8682213/v1/1ac819a1d575896ec489f39e.png"},{"id":102852280,"identity":"6fbfccfc-3ca0-4a5e-bc82-e90d629c3e40","added_by":"auto","created_at":"2026-02-17 14:36:06","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1159451,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCase studies of good and poor outcome patients. a \u003c/strong\u003ePredicted epileptogenic zone (PEZ) from spike (SF, purple), ripple (RF, blue), and combined (CF, red) feature-based Random Forest classifiers in two good outcome patients (#6 and #19) and two poor outcome patients (#37 and #45). Green areas in the figures represent resection and white circles represent contacts.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-8682213/v1/5bbb0b1da917ac31bbb3de25.png"},{"id":105755234,"identity":"75efc908-81a7-4937-b7d2-034e269a64cb","added_by":"auto","created_at":"2026-03-30 16:26:21","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4723389,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8682213/v1/70074b0c-77dd-49e1-9229-854347fd95ec.pdf"},{"id":102852282,"identity":"3637bc79-29c7-4c38-90f3-b63c68aaa8eb","added_by":"auto","created_at":"2026-02-17 14:36:06","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":2354593,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-8682213/v1/8ceae4109f091f6d0fdaa151.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Predicting Surgical Outcome in Drug-Resistant Epilepsy by Combining Interictal Biomarkers within a Machine Learning Framework","fulltext":[{"header":"Introduction","content":"\u003cp\u003ePrecise delineation of the epileptogenic zone (EZ) remains one of the most significant challenges for surgery planning and achieving seizure freedom post-surgery in patients with drug-resistant epilepsy (DRE).\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e This hypothetical zone is often approximated by the seizure onset zone (SOZ), which is identified by manual review of ictal activity captured by intracranial EEG (iEEG) to detect the brain region(s) that initiate clinical seizures.\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e Yet, increasing evidence suggests that interictal electrophysiological abnormalities observed in iEEG recordings (i.e., brief pathological events occurring between seizures) may also approximate the EZ.\u003csup\u003e3,4\u003c/sup\u003e Importantly, identifying such biomarkers from readily available interictal data offers a significant advantage by eliminating the need to wait for a seizure to occur.\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e,\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eAmong the proposed interictal epilepsy biomarkers, spikes and high-frequency oscillations (HFOs) are the most extensively studied.\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e Interictal spikes, brief high-amplitude deflections typically occurring in the 1\u0026ndash;70 Hz frequency band, are routinely used in clinical practice as markers of the irritative zone.\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e,\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e HFOs, which are considered more specific biomarkers of the EZ compared to spikes\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e,\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e, are commonly classified into ripples (80\u0026ndash;250 Hz) and fast ripples (250\u0026ndash;500 Hz).\u003csup\u003e\u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e Fast ripples, although closely linked to the EZ\u003csup\u003e15\u0026ndash;17\u003c/sup\u003e, are less frequently observed in patients with DRE, and their low sensitivity may limit their value for the EZ localization.\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e,\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e On the other hand, ripples are present in most patients with DRE but are also observed in physiological brain regions.\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e,\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e Thus, the presurgical value of interictal biomarkers, such as the spikes and HFOs, as biomarkers of the EZ remains largely debated since the cortical areas that generate them often extend beyond the EZ into brain regions that do not need to be resected or ablated for the patient to become seizure-free.\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e,\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eBecause of these limitations, temporal (i.e., timing and temporal occurrence), spectral (i.e., frequency content), and spatial (i.e., distribution across brain regions) features of spikes and ripples have been examined as potential indicators of epileptogenicity. In the temporal domain, the spike and ripple rates have been linked to the underlying brain tissue\u0026rsquo;s epileptogenicity.\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e,\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e,\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e Yet, spike-generating regions correspond to the broader irritative zone, which can extend beyond the EZ and overlap with physiological brain regions.\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e,\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e Moreover, not all ripples are pathological, as physiological ripples also occur, particularly in brain regions such as the hippocampus, occipital cortex, and paracentral areas.\u003csup\u003e\u003cspan additionalcitationids=\"CR29\" citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e Consequently, spike and ripple rates alone are unreliable predictors of the EZ. Recent studies have shown that the temporal co-occurrence of spikes and ripples may predict the EZ more effectively, or with comparable efficacy, compared to spikes or ripples alone.\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e,\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e,\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e This improved performance likely reflects the co-occurrence of spikes and ripples as a more pathologic pattern of neuronal synchronization than either event in isolation.\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e Yet, the interpretation of ripples remains challenging due to their susceptibility to contamination by artifacts arising from the Gibbs phenomenon, particularly in recordings containing epileptiform spikes.\u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e,\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e In the spectral domain, frequency-based features have been shown to identify abnormal oscillatory activity associated with epileptogenic networks.\u003csup\u003e\u003cspan additionalcitationids=\"CR36\" citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e However, they are limited by their sensitivity to sensor noise and biological artifacts.\u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e,\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e Spatial analyses have demonstrated that spikes and ripples often originate in specific brain regions and propagate to adjacent and distant areas within a few tens of milliseconds.\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e,\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e,\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e Recent studies have also shown that the onset generator of spikes and ripples (i.e., regions where discharges show the earliest peaks) are correlated with the EZ,\u003csup\u003e10,40,42,43\u003c/sup\u003e often outperforming the clinically defined SOZ in predicting the surgical outcome.\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e,\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e However, the spatial distribution of spikes and ripples fluctuates over time, shifting among multiple regions rather than arising from a single consistent source.\u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e,\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e Therefore, no single biomarker fully defines the EZ; rather, each reflects only a partial and variable aspect of epileptogenicity, highlighting the need for complementary approaches.\u003c/p\u003e \u003cp\u003eIt is reasonable to presume that the integration of temporal, spectral, and spatial features of interictal biomarkers will yield a more comprehensive and precise delineation of the EZ.\u003csup\u003e47\u0026ndash;49\u003c/sup\u003e Yet, the practical implementation of this approach is technically challenging. iEEG recordings are often performed with several dozens of contacts sampled at high rates (\u0026gt;\u0026thinsp;1 kHz) and collected over several hours. As a result, they generate a massive amount of data, making the process time-consuming, labor-intensive, and prone to bias and error. This underscores the need to develop novel automated approaches that integrate the complementary strengths of these biomarkers and features into a unified framework, enabling a more objective, efficient, and accurate characterization of the EZ.\u003c/p\u003e \u003cp\u003eMachine learning (ML) offers a robust framework for automating the analysis of large-scale data and integrating heterogeneous, high-dimensional features that are difficult to analyze manually or with conventional statistical methods.\u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e,\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e Previous studies have incorporated temporal, spectral, and spatial features using ML to delineate the EZ in patients with DRE.\u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e,\u003cspan additionalcitationids=\"CR53 CR54 CR55 CR56 CR57 CR58\" citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e However, few studies have further compared the performance of individual features with multi-feature ML-based frameworks\u0026rsquo; findings.\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e,\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e This comparison is critical, as the epileptogenic brain regions identified by individual spike and ripple features may differ across patients, or map incompletely to the EZ or extend beyond it. These limitations highlight the need for methods that systematically integrate heterogeneous biomarkers for EZ prediction. ML-based frameworks are well-suited for this task, as they can fuse multiple feature types, thereby improving accuracy, robustness, and interpretability in EZ mapping and surgical outcome prediction.\u003c/p\u003e \u003cp\u003eIn this study, we propose an ML-based multimodal framework that integrates diverse interictal spike and ripple temporal, spectral, and spatial features to automatically identify the EZ and predict postsurgical outcome in patients with DRE. We hypothesize that the proposed framework outperforms single-feature analyses in delineating the EZ and predicting the outcome by capturing nonlinear and cross-domain relationships among epileptogenic biomarkers. To evaluate this framework, automatic spike and ripple detectors were applied to iEEG recordings from children and young adults with DRE to extract multidomain electrode-level features. These features were then used to train Random Forest classifiers, with different combinations of features as predictors and resection as the target, to define the predicted EZ (PEZ). The PEZs derived from individual spike and ripple features and from the proposed feature-combining approach were analyzed to evaluate their ability to predict the EZ. We then assessed whether resection of each PEZ could predict surgical outcome. By combining multimodal interictal features, this study establishes a data-driven framework for more accurate localization of the EZ and prediction of surgical outcome for patients with DRE.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003ePatient cohort\u003c/h2\u003e \u003cp\u003eWe retrospectively analyzed interictal iEEG data from 62 children and young adults (29 females) with focal onset DRE, having either a good (34 patients, Engel I) or poor (28 patients, Engel \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\u0026gt;\\)\u003c/span\u003e\u003c/span\u003eI) surgical outcome following epilepsy surgery [median follow-up: 2.5 years (2\u0026ndash;5)], who satisfied the study\u0026rsquo;s inclusion and exclusion criteria (see \u003cb\u003eMaterials and Methods\u003c/b\u003e). Patient demographics are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. The patients had a median age at surgery of 12 years (range: 8-15.5) with a median seizure onset age of 4.5 years (2\u0026ndash;9). Our dataset included patients with three implantation types: (i) subdural electrodes [electrocorticography (ECoG)] (13 patients; eight with good outcome); (ii) stereotactic EEG (sEEG) implantation (36 patients; 20 with good outcome); and (iii) both subdural and depth electrodes (13 patients; 6 with good outcome). The median number of implanted contacts was 124 (102\u0026ndash;151). Seventeen patients (27.4%) had non-lesional MRI, 34 patients (54.8%) had a malformation of cortical development, and 11 patients (17.8%) had acquired brain injury. Twenty-six patients (41.9%) had temporal lobe epilepsy. No significant differences in these characteristics were found between good and poor outcome patients (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePatient demographic information categorized by post-surgical outcome.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCharacteristic\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eGood outcome\u003c/p\u003e \u003cp\u003e(Engel\u0026thinsp;=\u0026thinsp;I)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePoor outcome\u003c/p\u003e \u003cp\u003e(Engel\u0026thinsp;\u0026gt;\u0026thinsp;I)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePatients\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e28\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eGender\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.17\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e31\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e31\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAge at surgery\u003c/b\u003e, yr., median (IQR).\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e12 (8-15.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12 (7-15.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e12 (8\u0026ndash;15)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.8 \u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAge at seizure onset\u003c/b\u003e, yr., median (IQR).\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4.5 (2\u0026ndash;9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6 (2-10.25)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4 (2-6.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.08\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eFollow up\u003c/b\u003e, yr., median (IQR).\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.5 (2\u0026ndash;5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.5 (2\u0026ndash;5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.5 (1\u0026ndash;5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.8\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eNumber of contacts\u003c/b\u003e, no., median (IQR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e124 (102\u0026ndash;151)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e112 (88\u0026ndash;136)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e138 (106\u0026ndash;168)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.09\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eType of implantations\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.73\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSubdural (ECoG)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003esEEG\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e36\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSubdural and depth\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eImplantation side\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.96\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eL\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e40\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eL\u0026thinsp;+\u0026thinsp;R\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eEplilepsy localization\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.16\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTemporal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e26\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eExtra-temporal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e36\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMRI Findings\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.73\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNL\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMCD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eACQ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"5\"\u003eACQ\u0026thinsp;=\u0026thinsp;Acquired (i.e., stroke, neoplasm, and traumatic brain injury); ECoG\u0026thinsp;=\u0026thinsp;Electrocorticography; IQR\u0026thinsp;=\u0026thinsp;Inter-Quartile Range; L\u0026thinsp;=\u0026thinsp;Left; MCD\u0026thinsp;=\u0026thinsp;Malformation of Cortical Development (i.e., focal cortical dysplasia, polymicrogyria, gliosis, and tuberous sclerosis complex); NL\u0026thinsp;=\u0026thinsp;Non-Lesional; no. = Number; \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;\u003cem\u003ep\u003c/em\u003e-values; R\u0026thinsp;=\u0026thinsp;Right; sEEG\u0026thinsp;=\u0026thinsp;stereotactic electroencephalography; yr. = Year. \u003csup\u003ea\u003c/sup\u003e \u003cem\u003eChi-squared\u003c/em\u003e test; \u003csup\u003eb\u003c/sup\u003e \u003cem\u003eWilcoxon rank-sum\u003c/em\u003e test (two-tailed).\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eOverall processing Pipeline\u003c/h3\u003e\n\u003cp\u003eThe proposed framework automatically identifies the EZ from interictal iEEG data through automated detection of spikes and ripples, spike and ripple features extraction, and ML (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). We initially preprocessed a 5-minute interictal iEEG segment to remove artifacts and noisy channels (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea). The data were then bandpass filtered into the spike (1\u0026ndash;70 Hz)\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e and ripple (80\u0026ndash;250 Hz)\u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e frequency bands. Spikes and ripples were detected using previously validated fully automated algorithms.\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e,\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e From the detected spikes and ripples, we extracted a set of temporal, spectral, and spatial features. For the delineation of the EZ, we used as gold standard the resected volume of patients with good surgical outcome: channels inside the resection were labeled as EZ, and those outside as non-EZ. We first evaluated the individual discriminatory power of each feature for classifying the EZ vs. non-EZ channels. We then trained Random Forest classifiers to automatically identify the predicted EZ (PEZ). Finally, we assessed whether resecting the PEZ, which was mapped by each feature and the classifiers, could predict surgical outcome (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo characterize the epileptogenic properties of each channel, we derived a set of temporal, spectral and spatial features from the detected spikes and ripples (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb-f). Using in-house algorithms, we first identified propagation sequences of both spike and ripple events (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb-c). For each type of event, we determined the onset and spread channels and derived multiple quantitative features per channel, including propagation indices (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb-c), power (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ed), spatial distribution (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ee), and spike and ripple rates (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ef). Finally, we computed the co-occurrence of spikes and ripples for each channel (see \u003cb\u003eMaterials and Methods\u003c/b\u003e). Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e summarizes the features adopted in this study. The selected features were designed to capture complementary temporal, spectral, and spatial characteristics of interictal spikes and ripples, known to reflect epileptogenic tissue. Spike and ripple rates (i.e. SR and RR, respectively) and power (i.e. SP and RP, respectively) quantify local excitability and pathological high-frequency activity\u003csup\u003e\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e,\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e\u003c/sup\u003e, while spike and ripple propagation indices (i.e. SPI and RPI, respectively) characterize onset and spread channels.\u003csup\u003e\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e The spatial distribution of interictal epileptiform events has been shown to reflect underlying epileptogenic organization and network spread, with events clustering around putative onset regions and dispersing along pathological propagation pathways.\u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e We used several distance-based measures to quantify the spatial distribution of epileptiform events relative to the contact with the highest spike and ripple rate (i.e. ED-HSR and ED-HRR, respectively), the highest spike and ripple power (i.e. ED-HSP and ED-HRP, respectively), and the spike and ripple onset (i.e. ED-SO and ED-RO, respectively) contacts. In addition, we included the spike\u0026ndash;ripple co-occurrence (SRC) to quantify pathological coupling between spikes and ripples, which reflects abnormal local network synchronization and has been strongly associated with epileptogenicity.\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e,\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e Together, these features integrate local and network-level information, providing a physiologically grounded and interpretable representation for automated delineation of the EZ.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSummary of spike and ripple features used for EZ characterization.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBiomarker\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFeature\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDescription\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eType\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNormalization\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003ePhysiological Meaning\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"5\" rowspan=\"6\"\u003e \u003cp\u003e\u003cb\u003eSpike\u003c/b\u003e\u003c/p\u003e \u003cp\u003e(1\u0026ndash;70 Hz)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eSR\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSpike rate (min⁻\u0026sup1;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTemporal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMin\u0026ndash;Max\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eInterictal spike excitability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eSP\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSpike power\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSpectral\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMin\u0026ndash;Max\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eSignal energy\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eSPI\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSpike propagation index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTemporal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMax-Min\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eDegree of spike spread across channels\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eED-HSR\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEuclidean distance from highest spike rate contact\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSpatial\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMax-Min\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eProximity to maximal spike-generating region\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eED-HSP\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEuclidean distance from highest spike power contact\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSpatial\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMax-Min\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eProximity to strongest spike source\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eED-SO\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEuclidean distance from spike onset contact\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSpatial\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMax-Min\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eDistance from earliest spike onset\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"5\" rowspan=\"6\"\u003e \u003cp\u003e\u003cb\u003eRipple\u003c/b\u003e\u003c/p\u003e \u003cp\u003e(80\u0026ndash;250 Hz)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eRR\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRipple rate (min⁻\u0026sup1;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTemporal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMin\u0026ndash;Max\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eHigh-frequency excitability\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eRP\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRipple power\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSpectral\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMin\u0026ndash;Max\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eSignal energy\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eRPI\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRipple propagation index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTemporal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMax-Min\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eSpatial spread of ripple activity\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eED-HRR\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEuclidean distance from highest ripple rate contact\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSpatial\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMax-Min\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eProximity to dominant ripple-generating region\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eED-HRP\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEuclidean distance from highest ripple power contact\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSpatial\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMax-Min\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eProximity to strongest ripple source\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eED-RO\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEuclidean distance from ripple onset contact\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSpatial\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMax-Min\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eDistance from earliest ripple onset\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSpike\u0026ndash;Ripple\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eSRC\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSpike\u0026ndash;ripple co-occurrence rate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTemporal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMin\u0026ndash;Max\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003ePathological coupling of spikes and ripples\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eWe next coregistered the pre- and post-operative MRIs with the electrode implantation computerized tomography (CT) to map the locations of iEEG electrodes and the resected brain regions (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eg). Pediatric epileptologists defined the SOZ through visual inspection of ictal iEEG data independently from this study. We finally trained Random Forest classifiers using different subsets of the extracted features as predictors: spike features alone (SF), ripple features alone (RF), and combined spike\u0026ndash;ripple features (CF). Resected contacts were labeled as epileptogenic, and all remaining contacts were considered non-epileptogenic (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eh). The contacts predicted as epileptogenic defined the PEZ. Training was performed only on good outcome patients, as their resections were assumed to contain the actual EZ, and model performance was evaluated using leave-one-patient-out cross-validation (LOPO-CV). We then quantified the PEZ with spatial measures: (i) focality (\u003cem\u003eF\u003c/em\u003e\u003csub\u003e\u003cem\u003ePEZ\u003c/em\u003e\u003c/sub\u003e); (ii) overlap with resection (\u003cem\u003eO\u003c/em\u003e\u003csub\u003e\u003cem\u003eRES\u003c/em\u003e\u003c/sub\u003e), and (iii) average distance from resection (\u003cem\u003eD\u003c/em\u003e\u003csub\u003e\u003cem\u003eRES\u003c/em\u003e\u003c/sub\u003e).\u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e We defined the \u003cem\u003eF\u003c/em\u003e\u003csub\u003e\u003cem\u003ePEZ\u003c/em\u003e\u003c/sub\u003e as the average Euclidean distance (in mm) between the PEZ contacts. We defined the \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\:\\)\u003c/span\u003e\u003c/span\u003eas the percentage of PEZ contacts within 10 mm of resection. We defined the \u003cem\u003eD\u003c/em\u003e\u003csub\u003e\u003cem\u003eRES\u003c/em\u003e\u003c/sub\u003e as the average Euclidean distance of the PEZ contacts to the resection. While training was performed using only good outcome patients, testing included both good and poor outcome patients. Finally, we tested whether the resection of the PEZ was associated with surgical outcome using the \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\:\\)\u003c/span\u003e\u003c/span\u003eas predictor. Additional analyses compared the performance of individual features and classifiers trained on different feature subsets in predicting the EZ and outcome. We also assessed how feature selection and implantation type impacted the accuracy of EZ and outcome prediction (see \u003cb\u003eMaterials and Methods\u003c/b\u003e).\u003c/p\u003e\n\u003ch3\u003ePredictive power of individual features\u003c/h3\u003e\n\u003cp\u003eTo assess the relationship between the extracted features and the EZ, we compared the normalized feature values inside vs. outside resection in good and poor outcome patients, separately. In good outcome patients, we observed increased SR (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), SP (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), SPI (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01), ED-HSR (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.0001), ED-HSP (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.0001), ED-SO (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.0001), RP (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), ED-HRR (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01), ED-HRP (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05), ED-RO (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and SRC (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) inside (compared to outside) resection (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea). In poor outcome patients, we observed higher SR (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05), SP (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05), ED-HSP (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05), ED-SO (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01), RP (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05), and ED-RO (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) inside (compared to outside) resection (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea). To assess the predictive power of each feature, we computed their performance in predicting the EZ in good outcome patients (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). First, we evaluated feature importance using one-way ANOVA by ranking the proportion of variance (η\u0026sup2;) explained by class labels (epileptogenic vs. non-epileptogenic) in decreasing order (see \u003cb\u003eMaterials and Methods\u003c/b\u003e). The top-ranked features ED-SO, ED-HSP, SR, ED-HSR, SPI, and ED-RO exhibited strong discriminative power in differentiating between epileptogenic and non-epileptogenic contacts (η\u0026sup2;\u0026gt;0.06) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). Next, we trained single-feature threshold-based classifiers and their performance in discriminating between epileptogenic and non-epileptogenic contacts using different performance metrics including sensitivity, specificity, precision, negative predictive value (NPV), accuracy, F1-measure, as well as the area under the receiver operating characteristic curve (AUC-ROC), precision\u0026ndash;recall curve (AUC-PR), and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e (see \u003cb\u003eMaterials and Methods\u003c/b\u003e). Among all features, SR, ED-HSR, ED-HSP, and ED-SO achieved the highest AUC-ROC and AUC-PR (\u0026gt;\u0026thinsp;74% and 66%, respectively) and exhibited\u0026thinsp;\u0026gt;\u0026thinsp;60% overlap with resection. The resection of the PEZ, estimated from most single-feature classifiers, also predicted outcome, except for the SPI, RR, and RPI (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003ePredictive power of feature sets\u003c/h3\u003e\n\u003cp\u003eThe performance of the Random Forest classifiers using different feature sets (i.e., SF, RF, CF) as predictors of the EZ was evaluated using the LOPO-CV, AUC-ROC, AUC-PR, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e computed for individual features and feature sets (see \u003cb\u003eMaterials and Methods\u003c/b\u003e). We found that SF and CF achieved higher AUC-ROC (0.9 and 0.89, respectively) compared to most individual features, except for the ED-SO (0.9). In terms of AUC-PR, SF, CF, and ED-SO showed higher values compared to all individual features (AUC-PR\u0026thinsp;=\u0026thinsp;0.76, 0.78, and 0.77, respectively). Many individual features, however, showed higher AUC-ROC and AUC-PR compared to RF. Regarding overlap with resection, SF, RF, and CF all achieved \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\:\\)\u003c/span\u003e\u003c/span\u003e\u0026gt; 74%, which was higher than the individual features (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\:\\)\u003c/span\u003e\u003c/span\u003e \u0026le; 70%) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eStatistical pairwise comparisons between the performance measures of individual features and classifiers showed that SF and CF outperformed all individual features in terms of AUC-ROC and AUC-PR, except for the ED-HSR, ED-HSP, and ED-SO. RF showed higher performance than some individual features but did not exceed SF or CF. No differences were observed between SF and CF for the AUC-ROC and AUC-PR. Regarding the \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e, SF and CF outperformed all individual features except the SRC, while no differences were observed among the SF, RF, and CF (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb).\u003c/p\u003e\n\u003ch3\u003eEffect of implantation type on EZ predictability\u003c/h3\u003e\n\u003cp\u003eTo determine whether the implantation type (i.e., ECoG, sEEG, or both subdural and depth electrodes) influenced the PEZ's predictive power, we trained Random Forest classifiers separately for cohorts sharing the same implantation type. For good outcome patients, we estimated the PEZ using SF, RF, and CF and then calculated the AUC-ROC, AUC-PR, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e (see \u003cb\u003eMaterials and Methods\u003c/b\u003e). We found no differences between the AUC-ROC, AUC-PR, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\:\\)\u003c/span\u003e\u003c/span\u003eof the PEZs with resection among the different implantation types for any of the feature sets (see \u003cb\u003eSupplementary Note 2\u003c/b\u003e).\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eFocality, proximity, and overlap of PEZ with resection\u003c/h2\u003e \u003cp\u003eTo assess the concordance of PEZ with RES, we compared \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{D}_{RES}\\)\u003c/span\u003e\u003c/span\u003e, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{F}_{PEZ}\\)\u003c/span\u003e\u003c/span\u003e across patients with good and poor outcomes for the three feature subsets. We found that the SF, RF, and CF had higher \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e and lower \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{D}_{RES}\\)\u003c/span\u003e\u003c/span\u003e in good compared to poor outcome patients (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01, \u003cem\u003ed\u003c/em\u003e \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\ge\\:0.4\\:\\)\u003c/span\u003e\u003c/span\u003e). No differences in \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{F}_{PEZ}\\)\u003c/span\u003e\u003c/span\u003e were observed between good and poor outcome patients for any of the feature sets. Notably in good outcome patients, the SF had a median \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\approx\\:\\)\u003c/span\u003e\u003c/span\u003e 74 (56\u0026ndash;95)%, RF had a median \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\approx\\:\\)\u003c/span\u003e\u003c/span\u003e 75 (50\u0026ndash;85)%, and CF had a median \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\approx\\:\\)\u003c/span\u003e\u003c/span\u003e 74 (57\u0026ndash;94)% (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eEZ and Surgical Outcome Prediction Across Features and Classifiers\u003c/h3\u003e\n\u003cp\u003eTo evaluate the PEZ as an interictal biomarker of the EZ, we examined its ability to predict surgical outcome (see \u003cb\u003eMaterials and Methods\u003c/b\u003e). We found that all PEZ derived from SF, RF, and CF predicted outcome (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01, \u003cem\u003eFisher Exact\u003c/em\u003e test), while the clinically defined SOZ did not predict outcome (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec). Additionally, the CF model predicted outcome with the highest sensitivity (88%), specificity (68%), precision (77%), NPV (83%), F1-score (82%), and accuracy (79%) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec). We finally performed a pairwise McNemar test on the confusion matrices of outcome predictions of individual features, feature sets (i.e., SF, RF, CF), and the SOZ, to determine whether their outcome predictions were significantly different (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec). We found that SR was better than RPI and the SOZ, and ED-SO was better than the SOZ (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05, \u003cem\u003eMcNemar\u003c/em\u003e test). SF showed better predictive power compared to the RPI and the SOZ (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05, \u003cem\u003eMcNemar\u003c/em\u003e test). Finally, CF had higher predictive power of surgical outcome than all individual features except the SR, ED-HSR, ED-HSP, ED-SO, ED-HRP, SF, and RF (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05, \u003cem\u003eMcNemar\u003c/em\u003e test).\u003c/p\u003e\n\u003ch3\u003eRepresentative cases\u003c/h3\u003e\n\u003cp\u003eTo illustrate how the PEZ relates to the EZ, we report findings from four representative cases: two patients with good and two patients with poor surgical outcome. For these patients, we depict the PEZ overlayed on their MRIs and the resection volume (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). For good outcome patient #6, the PEZ derived from the CF (\u003cem\u003eO\u003c/em\u003e\u003csub\u003e\u003cem\u003eRES\u003c/em\u003e\u003c/sub\u003e= 71%) and SF (\u003cem\u003eO\u003c/em\u003e\u003csub\u003e\u003cem\u003eRES\u003c/em\u003e\u003c/sub\u003e= 51%) showed greater overlap with the resection volume than that from the RF (\u003cem\u003eO\u003c/em\u003e\u003csub\u003e\u003cem\u003eRES\u003c/em\u003e\u003c/sub\u003e= 41%) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). The PEZs from the CF and SF were also more focal (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{F}_{PEZ}\\)\u003c/span\u003e\u003c/span\u003e=22 mm and 23 mm, respectively) and closer to the resection (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{D}_{RES}\\)\u003c/span\u003e\u003c/span\u003e=12 mm and 13 mm, respectively) compared to the RF (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{F}_{PEZ}\\)\u003c/span\u003e\u003c/span\u003e=32 mm, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{D}_{RES}\\)\u003c/span\u003e\u003c/span\u003e=19 mm). For good outcome patient #19, PEZs from all three classifiers (i.e., SF, RF, and CF) had high overlap with the resection (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e\u0026gt;85%), were focal (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{F}_{PEZ\\:}\\)\u003c/span\u003e\u003c/span\u003e\u0026lt;23 mm), and close to the resection (\u0026lt;4mm) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn contrast, for the poor outcome patient #37, the PEZs from all three classifiers showed low overlap with resection (\u003cem\u003eO\u003c/em\u003e\u003csub\u003e\u003cem\u003eRES\u003c/em\u003e\u003c/sub\u003e \u0026lt; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:34\\%\\)\u003c/span\u003e\u003c/span\u003e) and low focality (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{F}_{PEZ}\\:\\)\u003c/span\u003e\u003c/span\u003e\u0026gt;31mm). Finally, for poor outcome patient #44, although the PEZs were focal (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{F}_{PEZ}\\)\u003c/span\u003e\u003c/span\u003e\u0026lt;20mm), they showed a low overlap with the resection (\u003cem\u003eO\u003c/em\u003e\u003csub\u003e\u003cem\u003eRES\u003c/em\u003e\u003c/sub\u003e \u0026lt;\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:41\\%\\)\u003c/span\u003e\u003c/span\u003e) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we introduce a novel supervised ML-based framework that delineates the EZ and predicts the surgical outcome in patients with DRE by extracting and combining features of interictal biomarkers from iEEG data in a fully automated fashion. Our framework detects spikes and ripples, extracts relevant temporal, spectral, and spatial features, and employs Random Forest classifiers to identify the PEZ. We evaluated the performance of both the individual features and Random Forest classifiers trained on SF, RF, and CF with RES as target. Our findings indicate that the SF and CF classifiers not only outperformed the models based on individual feature sets but also yielded a focal PEZ with high overlap and proximity to the resection in patients with good surgical outcome. Moreover, resection of the PEZ, defined by the CF, achieved an accuracy of ~\u0026thinsp;80% in predicting surgical outcome, while resection of the clinically defined SOZ showed limited predictive value. This discrepancy likely reflects fundamental limitations of the SOZ concept in sEEG and depth-electrode-based evaluations, where seizure onset may be rapidly distributed across mesial and lateral temporal structures\u003csup\u003e\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e, or reflect only the early propagation rather than the full epileptogenic network.\u003csup\u003e\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e,\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u003c/sup\u003e In addition, the SOZ is constrained by electrode coverage of the EZ, which may incompletely capture deeper or spatially remote regions contributing to seizure generation. As a result, SOZ localization alone may fail to identify all tissue necessary for resection and achievement of seizure freedom. Our framework generalizes across different implantation types and enables fully automated epileptogenic zone localization from interictal iEEG data, substantially reducing the need for manual identification and evaluation of epileptogenic brain regions.\u003c/p\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eSpike spatial distribution plays a key role in defining the EZ\u003c/h2\u003e \u003cp\u003eInterictal biomarkers such as spikes and ripples, as well as their features including event rate, power, spike-ripple co-occurrence and their onsets, have been previously associated with the EZ.\u003csup\u003e11,58,69\u0026ndash;72\u003c/sup\u003e Our results were consistent with these previous findings where several spike and ripple features had higher values within the resection (compared to outside) for good-outcome patients (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea). Notably, some features also showed elevated values within resection for patients with poor outcome but with lower effect size (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). These findings indicate that although each feature independently captures distinct aspects of the EZ, resection of brain regions based on individual features alone does not necessarily lead to seizure freedom. Upon ranking feature importance, distance-based spike features emerged as the most informative (showing higher effect sizes) than other features in distinguishing epileptogenic from non-epileptogenic contacts (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). This notion was further validated by the EZ prediction analysis, where performance metrics showed that the ED-SO, ED-HSR, and ED-HSP achieved higher AUC-ROC, AUC-PR, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\:\\)\u003c/span\u003e\u003c/span\u003evalues compared to the other features (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). This aligns with a recent study showing that the spatial distribution of biomarkers plays a key role in defining the EZ.\u003csup\u003e48\u003c/sup\u003e\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eCombined Features define the EZ better than individual features\u003c/h2\u003e \u003cp\u003eWhile each feature captures a distinct characteristic of the epileptogenic tissue, no single measure fully represents the complex spatial, spectral, and temporal dynamics of the EZ. A substantial body of literature suggests that ML-based integration of multiple electrophysiological biomarkers yields a more accurate localization of the EZ compared to individual features alone.\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e,\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e,\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e,\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e,\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e,\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e\u003c/sup\u003e Such integration facilitates the discrimination between epileptogenic and non-epileptogenic contacts, even when measurements are incomplete or noisy. Our findings support this notion, as SF and CF classifiers outperformed individual features when predicting the epileptogenic and non-epileptogenic contacts. Specifically, SF and CF outperformed all individual features in terms of AUC-ROC and AUC-PR (except for the ED-HSR, ED-HSP, and ED-SO) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea). Moreover, SF and CF exhibited higher \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e than all individual features, except for the SRC (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb). Overall, our results highlight the added value of integrating multiple spatial, temporal, and spectral interictal features to improve the delineation of PEZ and support the notion that ML classifiers can capture nonlinear feature relationships of different epileptogenic activity more effectively than single-feature approaches.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eRipple features provide modest improvement over spike features\u003c/h2\u003e \u003cp\u003eThe Random Forest classifiers trained using the CF slightly outperformed those trained on SF, suggesting that spike features capture most of the relevant epileptogenic activity (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). While combining spike and ripple features provided a more accurate representation of the PEZ than any individual feature alone, our results indicate that ripple features provided only a modest improvement over spike-based features. This may be due to the fact that not all ripples are pathological, underscoring the importance of distinguishing pathological from physiological ripples, which also occur in non-epileptogenic brain regions and during normal brain fucntion.\u003csup\u003e\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e,\u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e\u003c/sup\u003e Although efforts have been made to address this distinction suggesting that pathological HFOs tend to co-occur with spikes and exhibit shorter durations and lower power and amplitude\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e,\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e,\u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e77\u003c/span\u003e,\u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e\u003c/sup\u003e, other studies reported the opposing findings.\u003csup\u003e\u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e79\u003c/span\u003e,\u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e80\u003c/span\u003e\u003c/sup\u003e Collectively, our findings suggest that, although integrating multiple biomarkers can offer incremental gains, spike features remain the dominant contributor to accurate identification of the PEZ based on interictal data. These findings are consistent with a recent study showing that low-frequency features alone were sufficient for the accurate delineation of the EZ and prediction of surgical outcome, with high-frequency features providing little or no additional value.\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eCombined Features predict outcome with higher accuracy than individual features\u003c/h2\u003e \u003cp\u003ePrevious studies have reported strong associations between interictal biomarkers and postsurgical outcome.\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e,\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e,\u003cspan additionalcitationids=\"CR82 CR83 CR84 CR85\" citationid=\"CR81\" class=\"CitationRef\"\u003e81\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR86\" class=\"CitationRef\"\u003e86\u003c/span\u003e\u003c/sup\u003e These results indicate that, in patients with focal epilepsy, seizure-generating regions form a distinct pathological network isolated from surrounding brain tissue. When these pathological regions are completely resected, patients are more likely to achieve postoperative seizure freedom. Indeed, the spatial concordance between the identified epileptogenic regions and the resected zone has previously been used as a reliable predictor of success.\u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e,\u003cspan citationid=\"CR82\" class=\"CitationRef\"\u003e82\u003c/span\u003e,\u003cspan citationid=\"CR85\" class=\"CitationRef\"\u003e85\u003c/span\u003e,\u003cspan citationid=\"CR87\" class=\"CitationRef\"\u003e87\u003c/span\u003e\u003c/sup\u003e Our findings support this notion since we observed a decrease in the \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e as well as increased \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{D}_{RES}\\)\u003c/span\u003e\u003c/span\u003e in poor outcome patients (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). These findings may indicate the presence of other epileptogenic regions distant from the resection, which may not have been resected due to overlap of the EZ with eloquent areas or could have been entirely missed [e.g., patient #37 and #44 (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e)]. Upon assessing the ability of individual features to predict surgical outcome, we found that almost all features were predictive, except for the SPI, RR, and RPI (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). All Random Forest classifiers were predictive of the surgical outcome (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb). When comparing outcome predictive performance across all individual features and classifiers, CF outperformed many of the features but not SR, ED-HSR, ED-HSP, ED-SO, ED-HRP, SF, and RF. These findings suggest that outcome prediction remains feasible using individual features alone, and that combining features does not necessarily improve performance, as many individual features are already informative (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eLimitations\u003c/h2\u003e \u003cp\u003eDespite these promising results, several limitations should be acknowledged. Our retrospective study analyzed iEEG data from a single-center cohort of patients with DRE and heterogeneous pathologies. Because the models were trained using data from good outcome patients, future studies from larger, multicenter cohorts are needed to evaluate its clinical applicability, particularly within more homogeneous patient populations.\u003c/p\u003e \u003cp\u003eResection volumes from good outcome patients are often used as a gold standard for the EZ, despite the possibility that they include non-epileptogenic tissue.\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR88\" class=\"CitationRef\"\u003e88\u003c/span\u003e\u003c/sup\u003e This may result in either missing critical epileptogenic regions or defining areas larger than the actual EZ. Because ML model performance depends strongly on the accuracy of the ground truth used, improved approximations of the EZ are needed to provide more reliable targets for ML training. Moreover, iEEG is spatially limited since it is based on a subjective hypothesis agreed upon during presurgical evaluation. Consequently, the EZ was likely sampled well in good outcome patients but may have been missed in poor outcome patients. Advances in whole-brain iEEG implantations may overcome these drawbacks. Furthermore, future studies on longer data segments may be conducted to enhance the accuracy of PEZ in delineating the EZ. Finally, the retrospective nature of the study warrants further validation in larger, prospective cohorts to confirm the robustness and generalizability of these findings.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn this study, we systematically evaluated the predictive power of individual features and combined feature-based classifiers for localizing the EZ and predicting surgical outcome in patients with DRE. Our results demonstrate that Random Forest classifiers outperform individual features, providing more focal and accurate identification of the EZ. While spike features remain highly informative for outcome prediction, the integration of ripple features offers a modest additional advantage. Importantly, these findings support the use of short segments of interictal data to approximate the EZ and inform resection planning, potentially reducing reliance on ictal recordings. Future prospective studies with larger cohorts are necessary to validate and refine this approach for clinical translation.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003ePatient Cohort\u003c/h2\u003e \u003cp\u003eWe retrospectively analyzed iEEG data of patients with DRE, from Boston Children\u0026rsquo;s Hospital (BCH, 51 patients, admitted between June 2011 and June 2023) and Cook Children\u0026rsquo;s Medical Center (CCMC, 11 patients, admitted between December 2019 and August 2021). The patients had long-term monitoring with iEEG and underwent resection neurosurgery. The selection of the patients was based on the following criteria: (i) availability of at least 5-minute interictal iEEG epochs with no biological or technical artifacts; (ii) availability of post-implantation computerized tomography (CT); (iii) availability of pre-operative and post-operative MRIs; (iv) accurate information about the resection volume, resected contacts, and the clinically defined SOZ; (v) availability of post-surgical outcome at least one year after surgery; (vi) data containing spikes and ripples; (vii) sampling frequency\u0026thinsp;\u0026gt;\u0026thinsp;1,000 Hz. The surgical outcome was determined by a pediatric epileptologist after multiple follow-up visits. We used the Engel score\u003csup\u003e\u003cspan citationid=\"CR89\" class=\"CitationRef\"\u003e89\u003c/span\u003e\u003c/sup\u003e to classify patients as good (Engel 1) or poor outcome (Engel 2\u0026ndash;4). The patient demographic and the clinical information of the cohort are provided in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. The protocol was approved by North Texas Regional IRB (2019\u0026thinsp;\u0026minus;\u0026thinsp;166; PI: C. Papadelis), that waived the need for informed consent considering the study\u0026rsquo;s retrospective nature. All methods and analyses were performed in accordance with relevant guidelines and regulations.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eInterictal iEEG recordings\u003c/h2\u003e \u003cp\u003eAt BCH, long-term iEEG data were collected using stereotactic electrodes, subdural grids, or combined subdural grids and depth electrodes utilizing XLTEK Quantum NeuroWorks (Natus Inc., USA). The subdural electrodes had a diameter of 2\u0026ndash;3 mm with a 10 mm inter-contact spacing, while the depth electrodes consisted of 6 to 16 contacts with a diameter of 0.8 mm and were linearly arranged approximately 3\u0026ndash;5 mm apart. At CCMC, long-term iEEG data were collected using stereotactic electrodes utilizing Quantum NeuroWorks (Natus Inc., USA). The electrodes consisted of 8 to 16 contacts with a diameter of 0.8 mm and were linearly arranged approximately 3\u0026ndash;5 mm apart.\u003c/p\u003e \u003cp\u003eThe data were recorded with a sampling frequency of 1,000 to 2,048 Hz. We specifically chose 5-minute interictal segments during non-REM slow-wave sleep (when applicable), occurring at least one hour before or after clinical seizures or half an hour before or after an electrographic seizure.\u003csup\u003e\u003cspan citationid=\"CR90\" class=\"CitationRef\"\u003e90\u003c/span\u003e\u003c/sup\u003e This approach ensured the incorporation of epochs with the highest spike rate and minimal motion artifacts.\u003csup\u003e\u003cspan citationid=\"CR91\" class=\"CitationRef\"\u003e91\u003c/span\u003e,\u003cspan citationid=\"CR92\" class=\"CitationRef\"\u003e92\u003c/span\u003e\u003c/sup\u003e Noisy channels and short segments with artifacts were excluded.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003eStructural Imaging and iEEG recordings\u003c/h2\u003e \u003cp\u003eMRI scans utilizing standard magnetization-prepared rapid acquisition gradient-echo sequences were performed both before and following resection with a 3T scanner (TIM TRIO, Siemens AG at BCH and Magnetom Skyra at CCMC). Post-implantation CT scans, with voxel size set at 0.5 x 0.5 x 0.5 mm\u0026sup3;, were performed after iEEG electrode implantation to document electrode locations. We aligned the electrode coordinates by registering the post-implantation CT scans with the preoperative MRI using \u003cem\u003eBrainstorm\u003c/em\u003e.\u003csup\u003e\u003cspan citationid=\"CR93\" class=\"CitationRef\"\u003e93\u003c/span\u003e\u003c/sup\u003e We then projected the electrode locations onto the patient's cortical surface reconstructed from their preoperative MRI with \u003cem\u003eFreesurfer\u003c/em\u003e.\u003csup\u003e\u003cspan citationid=\"CR94\" class=\"CitationRef\"\u003e94\u003c/span\u003e\u003c/sup\u003e We further adjusted the stereotactic and depth electrode positions to accommodate for brain shifting and pneumocephalus post-electrocorticography implantation.\u003csup\u003e\u003cspan citationid=\"CR95\" class=\"CitationRef\"\u003e95\u003c/span\u003e\u003c/sup\u003e The number and type of implanted electrodes for each patient are reported in \u003cb\u003eSupplementary Tables\u0026nbsp;1 and 2\u003c/b\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003eDefining the SOZ and resection contacts\u003c/h2\u003e \u003cp\u003eFor each patient, pediatric epileptologists determined the SOZ by visually examining ictal data. iEEG contacts that exhibited altered dynamics at the onset of seizures were designated as SOZ contacts. This process was conducted independently of the current study. To delineate the resection volume, we registered the preoperative and postoperative MRIs and manually outlined the boundaries of the resection volume on sequential slices using \u003cem\u003eBrainstorm\u003c/em\u003e\u003csup\u003e\u003cspan citationid=\"CR93\" class=\"CitationRef\"\u003e93\u003c/span\u003e\u003c/sup\u003e and MATLAB\u0026rsquo;s \u003cem\u003evolumeSegmentor\u003c/em\u003e toolbox. Subsequently, contacts\u0026thinsp;\u0026le;\u0026thinsp;10 mm from resection were considered as resected (RES).\u003c/p\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003eSpike and Ripple Features\u003c/h2\u003e \u003cp\u003eTo generate the features used in our classifiers, we focused on quantifying the spectral and spatiotemporal characteristics of spike and ripple propagations. We developed an in-house MATLAB algorithm to identify propagation sequences of spikes and ripples separately across all iEEG contacts for each patient.\u003csup\u003e\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e After detecting the spikes, we employed our algorithm that marks the earliest spike within each sequence as onset (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb\u003cb\u003e)\u003c/b\u003e.\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e Then, it adds the next spike to the propagation sequence if it occurred within a predefined time window, i.e. the inter-event latency. The contacts in every propagation were indexed sequentially, starting from one, denoting the onset contact, followed by increasing numbers for the subsequent spikes. A similar approach was adopted for mapping ripple propagations (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ec). Previous studies have shown that the inter-event latency for spikes and ripples is in the range of 3\u0026ndash;15 ms\u003csup\u003e10,96\u003c/sup\u003e and 10\u0026ndash;32 ms\u003csup\u003e10,42\u003c/sup\u003e, respectively. Based on these findings, we set here the inter-event latency to 10 and 30 ms, to map spike and ripple propagations respectively.\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e We also discarded propagations if more than 50% of their events occurred within 2 ms from each other.\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e,\u003cspan citationid=\"CR97\" class=\"CitationRef\"\u003e97\u003c/span\u003e\u003c/sup\u003e Using spikes and their propagation events, we computed the following features for each channel (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ed-f): (i) spike rate (SR); (ii) average spike power (SP); (iii) median spike propagation index (SPI); (iv) average Euclidean distance of contacts from the contact with the highest spike rate (ED-HSR); (v) average Euclidean distance of contacts from the contact with the highest average spike power (ED-HSP); and (vi) average Euclidean distance of the contacts from spike onset contact (ED-SO). Similarly, using ripples and their propagation events, we computed for each channel the: (i) ripple rate (RR); (ii) average ripple power (RP); (iii) median ripple propagation index (RPI); (iv) average Euclidean distance of contacts from the contact with the highest ripple rate (ED-HRR); (v) average Euclidean distance of contacts from the contact with the highest ripple power (ED-HRP); and (vi) average Euclidean distance of the contacts from ripple onset contact (ED-RO). Finally, we identified the cooccurrence of spikes and ripples in each contact and computed the average spike-ripple cooccurrence rate (SRC) per channel. Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e summarizes the features adopted in this study.\u003c/p\u003e \u003cp\u003eWe normalized the SR, RR, SP, RP and SRC between 0 and 1 using Min-Max scaling ensuring that the minimum value is scaled to 0 and the maximum to 1. For a feature \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:x\\)\u003c/span\u003e\u003c/span\u003e, the min-max normalized value \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:x\u0026rsquo;\\:\\)\u003c/span\u003e\u003c/span\u003eis defined by:\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e\n$$\\:{x}^{{\\prime\\:}}=\\frac{x-{x}_{min}}{{x}_{max}-{x}_{min}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{x}_{min}\\)\u003c/span\u003e\u003c/span\u003e is the minimum value and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{x}_{max}\\)\u003c/span\u003e\u003c/span\u003e is the maximum.\u003c/p\u003e \u003cp\u003eThe SPI, RPI, ED-HSR, ED-HRR, ED-HSP, ED-HRP, ED-SO, and ED-RO were scaled using Max-Min scaling ensuring that the maximum value is scaled to 0 and the minimum to 1. For a feature \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:x\\)\u003c/span\u003e\u003c/span\u003e, the max-min normalized value \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:x\u0026rsquo;\\:\\)\u003c/span\u003e\u003c/span\u003eis defined by:\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e\n$$\\:x{\\prime\\:}=\\frac{{x}_{max}-x}{{x}_{max}-{x}_{min}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003ePerformance of individual features in EZ and outcome prediction\u003c/h2\u003e \u003cp\u003eFor each feature, we estimated and compared the normalized values inside and outside the SOZ and resection for good and poor outcome patients, separately (\u003cem\u003eWilcoxon signed-rank\u003c/em\u003e test).\u003c/p\u003e \u003cp\u003eThen, to further evaluate the predictive power of individual features, we used the normalized values of each feature as predictors and RES as target. We performed the analysis only for good outcome patients since their resected tissue is assumed to contain the actual EZ. First, we assessed feature importance using one-way ANOVA. For each feature, we quantified the proportion of variance explained by the class labels (η\u0026sup2;)\u003csup\u003e98\u003c/sup\u003e, which provides a measure of how strongly a feature differentiates between classes (i.e. epileptogenic vs. non-epileptogenic). Values of η\u0026sup2; greater than 0.06 generally indicate that a feature has a meaningful influence on distinguishing between epileptogenic and non-epileptogenic classes. Next, for each feature, we used the LOPO-CV approach to assess its ability to predict the RES. At each step, one patient was left out, and the remaining patients\u0026rsquo; feature values were collectively used to compute the optimal thresholds based on the maximum Youden index (J) obtained from the ROC curves. Then, the left-out patient\u0026rsquo;s features were thresholded using the computed optimal threshold to identify PEZ. We then defined the following: (i) true positives (TP), PEZ contacts that were located within the RES; (ii) false positives (FP), PEZ contacts that were located outside the RES; (iii) false negatives (FN), non-PEZ contacts that were located inside the RES; and (iv) true negatives (TN), non-PEZ contacts that were located outside the RES. We then calculated the following performance metrics: (i) sensitivity [TP/(TP\u0026thinsp;+\u0026thinsp;FN)]; (ii) specificity [TN/(TN\u0026thinsp;+\u0026thinsp;FP)]; (iii) precision [TP/(TP\u0026thinsp;+\u0026thinsp;FP)]; (iv) NPV [TN/(TN\u0026thinsp;+\u0026thinsp;FN)]; (v) accuracy [(TP\u0026thinsp;+\u0026thinsp;TN)/(TP\u0026thinsp;+\u0026thinsp;TN\u0026thinsp;+\u0026thinsp;FP\u0026thinsp;+\u0026thinsp;FN)]; (vi) F1-score [2\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\times\\:\\)\u003c/span\u003e\u003c/span\u003eTP/(2\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\times\\:\\)\u003c/span\u003e\u003c/span\u003eTP\u0026thinsp;+\u0026thinsp;FP\u0026thinsp;+\u0026thinsp;FN)]; (vii) AUC-ROC; (viii) AUC-PR. The overlap with resection (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES})\\)\u003c/span\u003e\u003c/span\u003e was defined as the percentage of PEZ contacts within a 10 mm distance from the resection. The selection of the 10 mm cut-off was based on studies that showed that the gyral width is between 11\u0026ndash;21 mm.\u003csup\u003e99\u003c/sup\u003e We hypothesized that the resection of PEZ is linked to favorable surgical outcome, thus, we considered \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e as predictor and the dichotomized patient\u0026rsquo;s outcome as target. We considered: (i) TP, the number of good outcome patients who had\u0026thinsp;\u0026ge;\u0026thinsp;50% of the PEZ resected (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\:\\)\u003c/span\u003e\u003c/span\u003e\u0026ge; 50%); (ii) FN, the number of good outcome patients who had\u0026thinsp;\u0026lt;\u0026thinsp;50% of the PEZ resected (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\:\\)\u003c/span\u003e\u003c/span\u003e\u0026lt; 50%); (iii) FP, the number of poor outcome patients who had\u0026thinsp;\u0026ge;\u0026thinsp;50% of the PEZ resected; (iv) TN, the number of poor outcome patients who had\u0026thinsp;\u0026lt;\u0026thinsp;50% of the PEZ resected. We then constructed the confusion matrices for each feature and computed the \u003cem\u003eFisher\u0026rsquo;s exact\u003c/em\u003e test \u003cem\u003ep\u003c/em\u003e-values to find significance.\u003c/p\u003e \u003cdiv id=\"Sec25\" class=\"Section3\"\u003e \u003ch2\u003eTraining and evaluation of Random Forest classifiers\u003c/h2\u003e \u003cp\u003eWe trained and compared the performance of different Random Forest classifiers using different subsets of predictors and targets. We considered three sets of features as predictors: spike-features only (SF), ripple-features only (RF), and combined features (CF). We used RES to approximate the EZ. The setup resulted in training and evaluating three different classifiers with different combination of predictors (i.e., SF, RF, and CF). We exclusively trained and evaluated our classifiers using only patients with good outcome, since these patients remained seizure-free post-surgery, indicating successful removal of the EZ. Regarding the Random Forest classifier\u0026rsquo;s number of trees parameter, we conducted a five-fold cross-validation, varying the number of trees from 10 to 500 and selecting the value that yielded a stable AUC-ROC performance. Based on this analysis, we set the number of trees to 100 (\u003cb\u003eSupplementary Note 2\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eClassifier performance was evaluated using TP, TN, FP, and FN definitions identical to those used for individual feature analyses. Subsequently, we trained and tested the performance of the Random Forest classifier using LOPO-CV and computed performance metrics (i.e., sensitivity, specificity, precision, NPV, accuracy, F1-score, AUC-ROC, and AUC-PR). After estimating the PEZ using the trained classifiers, we computed \u003cem\u003eF\u003c/em\u003e\u003csub\u003e\u003cem\u003ePEZ\u003c/em\u003e\u003c/sub\u003e, \u003cem\u003eO\u003c/em\u003e\u003csub\u003e\u003cem\u003eRES\u003c/em\u003e\u003c/sub\u003e, and \u003cem\u003eD\u003c/em\u003e\u003csub\u003e\u003cem\u003eRES\u003c/em\u003e\u003c/sub\u003e. While training was performed using only good outcome patients, testing included both good and poor outcome patients.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section3\"\u003e \u003ch2\u003eRobustness of PEZ across implantation types\u003c/h2\u003e \u003cp\u003eTo evaluate whether our framework provides consistent findings among different implantation types, we examined the ability of PEZ to predict the EZ in good outcome patients with different implantation types. We first divided our good outcome patients into three subsets; (i) ECoG, patients with subdural implantations; (ii) sEEG, patients with stereotactic implantations; and (iii) ECoG and depth, patients with ECoG and a few depth electrodes (\u003cb\u003eSupplementary Table\u0026nbsp;1\u003c/b\u003e). For each subset, we trained and tested using CF with LOPO-CV and computed performance metrics (sensitivity, specificity, precision, NPV, accuracy, F1-score, AUC-ROC, AUC-PR) along with spatial properties (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e, Dice score, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{D}_{RES}\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{F}_{PEZ}\\)\u003c/span\u003e\u003c/span\u003e) of the PEZ.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section3\"\u003e \u003ch2\u003eRandom forest classifiers performance to predict surgical outcome\u003c/h2\u003e \u003cp\u003eWe presumed that the \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e of PEZ predicted from the classifiers SF, RF, and CF could predict outcome. Specifically, we assumed that in patients with good outcome, the PEZ would be located within resection, whereas in those with poor outcome, the resection would have missed the EZ. Subsequently, we performed a prediction analysis where we used the \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e of the PEZ as predictor of outcome. We considered good and poor outcomes as positive and negative classes, respectively. We then defined: (i) TP, the number of good outcome patients who had\u0026thinsp;\u0026ge;\u0026thinsp;50% of the PEZ resected (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e \u0026ge; 50%) ; (ii) FN, the number of good outcome patients who had\u0026thinsp;\u0026lt;\u0026thinsp;50% of the PEZ resected (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e \u0026lt; 50%); (iii) FP, the number of poor outcome patients who had\u0026thinsp;\u0026ge;\u0026thinsp;50% of the PEZ resected; and (iv) TN, the number of poor outcome patients who had\u0026thinsp;\u0026lt;\u0026thinsp;50% of the PEZ resected. We constructed confusion matrices for each feature set and computed the Fisher\u0026rsquo;s exact test \u003cem\u003eP\u003c/em\u003e-values to assess significance. Finally, we computed the sensitivity, specificity, precision, NPV, F1-score, and accuracy of the outcome predictions.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eThe Kolmogorov-Smirnov test was used to test the normality of features. \u003cem\u003eCliff\u0026rsquo;s d\u003c/em\u003e measure was used to compute effect sizes. Demographic data and PEZ spatial measures (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{F}_{PEZ}\\)\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e, Dice score, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{D}_{RES}\\)\u003c/span\u003e\u003c/span\u003e) are reported as median (25th -75th percentile). The \u003cem\u003echi-squared\u003c/em\u003e test was used to compare the effect of different cohort characteristics (i.e., gender, implantation side, epilepsy localization, and MRI findings) on outcome. We applied two-sided non-parametric \u003cem\u003eWilcoxon signed-rank\u003c/em\u003e test for all paired comparisons (median feature values inside vs. outside the SOZ and resection). We applied the two-sided \u003cem\u003eWilcoxon rank-sum\u003c/em\u003e test for non-paired comparisons between good and poor outcome patients. Bonferroni correction was applied in all multiple comparison tests.\u003c/p\u003e \u003cp\u003eWe used the one-sided \u003cem\u003eFisher\u0026rsquo;s exact\u003c/em\u003e test to evaluate the predictive value of the PEZs to predict outcome. To evaluate the performance of all individual features and trained Random Forest classifier models, we performed a comparative analysis of EZ prediction using the AUC-ROC, AUC-PR, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{O}_{RES}\\)\u003c/span\u003e\u003c/span\u003e. Pairwise comparisons were performed using the \u003cem\u003eWilcoxon signed-rank\u003c/em\u003e test on each performance metric, trained and evaluated using the LOPO-CV approach. False discovery Rate (FDR) was applied to correct for multiple comparisons. The results are displayed in a square matrix where each element indicates whether a significant difference (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) between the different predictor models was observed and, if so, identifies the model with the higher median value.\u003c/p\u003e \u003cp\u003eFor outcome prediction using individual features, classifier models, and SOZ, we performed a pairwise comparison using the \u003cem\u003eMcNemar\u003c/em\u003e test. The elements in the corresponding square matrix show both the presence of significant differences (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05) and the ID of the model with higher accuracy. The McNemar test was used to examine differences in the ability to predict outcomes using confusion matrices between different features and the proposed ML classifiers. All analysis was performed with MATLAB 2023b (The MathWorks, Inc.).\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research was supported by the National Institute of Neurological Disorders \u0026amp; Stroke (R01NS104116-01A1 and R01NS134944, PI: C. Papadelis).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data are available from the corresponding author upon request.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe code is available from the corresponding author upon request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eH.P., S.J., and C.P. conceptualized the project and experimental design. H.P. and S.J. developed the mathematical formulation and the codes of the proposed framework. H.P., S.J., S.P., E.T., J.M., J.B., S.S., P.P., and C.P., contributed to the acquisition and analysis of data. H.P. and S.J. participated in the development of various mathematical formulations, statistical analysis, and representation of results of the study. H.P., S.J., and C.P. contributed to drafting the text and preparing figures. All authors have read and approved the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors have no competing interests to disclose.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eJehi, L. The Epileptogenic Zone: Concept and Definition. \u003cem\u003eEpilepsy Curr.\u003c/em\u003e \u003cb\u003e18\u003c/b\u003e, 12\u0026ndash;16 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRosenow, F. \u0026amp; L\u0026uuml;ders, H. Presurgical evaluation of epilepsy. \u003cem\u003eBrain\u003c/em\u003e \u003cb\u003e124\u003c/b\u003e, 1683\u0026ndash;1700 (2001).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLai, N., Li, Z., Xu, C., Wang, Y. \u0026amp; Chen, Z. Diverse nature of interictal oscillations: EEG-based biomarkers in epilepsy. \u003cem\u003eNeurobiol. Dis.\u003c/em\u003e \u003cb\u003e177\u003c/b\u003e, 105999 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEmmady, P. D., Asuncion, R. M. D. \u0026amp; Anilkumar, A. C. \u003cem\u003eEEG Abnormal Waveforms. in StatPearls\u003c/em\u003e (StatPearls Publishing, 2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWidjaja, E. et al. Cost-effectiveness of pediatric epilepsy surgery compared to medical treatment in children with intractable epilepsy. \u003cem\u003eEpilepsy Res.\u003c/em\u003e \u003cb\u003e94\u003c/b\u003e, 61\u0026ndash;68 (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePapadelis, C. \u0026amp; Perry, M. S. Localizing the Epileptogenic Zone with Novel Biomarkers. \u003cem\u003eSemin Pediatr. Neurol.\u003c/em\u003e \u003cb\u003e39\u003c/b\u003e, 100919 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEngel, J. J. et al. Epilepsy biomarkers. \u003cem\u003eEpilepsia\u003c/em\u003e \u003cb\u003e54\u003c/b\u003e (Suppl 4), 61\u0026ndash;69 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStaley, K. J., White, A. \u0026amp; Dudek, F. E. Interictal spikes: harbingers or causes of epilepsy? \u003cem\u003eNeurosci. Lett.\u003c/em\u003e \u003cb\u003e497\u003c/b\u003e, 247\u0026ndash;250 (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStaley, K. J. \u0026amp; Dudek, F. E. Interictal spikes and epileptogenesis. \u003cem\u003eEpilepsy Curr.\u003c/em\u003e \u003cb\u003e6\u003c/b\u003e, 199\u0026ndash;202 (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTamilia, E. et al. Surgical resection of ripple onset predicts outcome in pediatric epilepsy. \u003cem\u003eAnn. Neurol.\u003c/em\u003e \u003cb\u003e84\u003c/b\u003e, 331\u0026ndash;346 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJahromi, S. et al. Overlap of spike and ripple propagation onset predicts surgical outcome in epilepsy. \u003cem\u003eAnn. Clin. Transl Neurol.\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e, 2530\u0026ndash;2547 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKlotz, K. A., Sag, Y., Sch\u0026ouml;nberger, J. \u0026amp; Jacobs, J. Scalp Ripples Can Predict Development of Epilepsy After First Unprovoked Seizure in Childhood. \u003cem\u003eAnn. Neurol.\u003c/em\u003e \u003cb\u003e89\u003c/b\u003e, 134\u0026ndash;142 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStaba, R. J., Wilson, C. L., Bragin, A., Fried, I. \u0026amp; Engel, J. J. Quantitative analysis of high-frequency oscillations (80\u0026ndash;500 Hz) recorded in human epileptic hippocampus and entorhinal cortex. \u003cem\u003eJ. Neurophysiol.\u003c/em\u003e \u003cb\u003e88\u003c/b\u003e, 1743\u0026ndash;1752 (2002).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJacobs, J. et al. Interictal high-frequency oscillations (80\u0026ndash;500 Hz) are an indicator of seizure onset areas independent of spikes in the human epileptic brain. \u003cem\u003eEpilepsia\u003c/em\u003e \u003cb\u003e49\u003c/b\u003e, 1893\u0026ndash;1907 (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eH\u0026ouml;ller, Y. et al. High-frequency oscillations in epilepsy and surgical outcome. A meta-analysis. \u003cem\u003eFront. Hum. Neurosci.\u003c/em\u003e \u003cb\u003e9\u003c/b\u003e, 574 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGonz\u0026aacute;lez Ot\u0026aacute;rula, K. A., von Ellenrieder, N., Cuello-Oderiz, C., Dubeau, F. \u0026amp; Gotman, J. High-Frequency Oscillation Networks and Surgical Outcome in Adult Focal Epilepsy. \u003cem\u003eAnn. Neurol.\u003c/em\u003e \u003cb\u003e85\u003c/b\u003e, 485\u0026ndash;494 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBragin, A., Engel, J. J., Wilson, C. L., Fried, I. \u0026amp; Mathern, G. W. Hippocampal and entorhinal cortex high-frequency oscillations (100\u0026ndash;500 Hz) in human epileptic brain and in kainic acid\u0026ndash;treated rats with chronic seizures. \u003cem\u003eEpilepsia\u003c/em\u003e \u003cb\u003e40\u003c/b\u003e, 127\u0026ndash;137 (1999).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWeiss, S. A. et al. Localizing epileptogenic regions using high-frequency oscillations and machine learning. \u003cem\u003eBiomark. Med.\u003c/em\u003e \u003cb\u003e13\u003c/b\u003e, 409\u0026ndash;418 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoehri, N. et al. High-frequency oscillations are not better biomarkers of epileptogenic tissues than spikes. \u003cem\u003eAnn. Neurol.\u003c/em\u003e \u003cb\u003e83\u003c/b\u003e, 84\u0026ndash;97 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, S. et al. Ripple classification helps to localize the seizure-onset zone in neocortical epilepsy. \u003cem\u003eEpilepsia\u003c/em\u003e \u003cb\u003e54\u003c/b\u003e, 370\u0026ndash;376 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMatsumoto, A. et al. Pathological and physiological high-frequency oscillations in focal human epilepsy. \u003cem\u003eJ. Neurophysiol.\u003c/em\u003e \u003cb\u003e110\u003c/b\u003e, 1958\u0026ndash;1964 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFabbri, L. et al. Spikes on ripples are better interictal biomarkers of epilepsy than spikes or ripples. \u003cem\u003eBrain Commun.\u003c/em\u003e \u003cb\u003e7\u003c/b\u003e, fcaf056 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZijlmans, M. et al. Ictal and interictal high frequency oscillations in patients with focal epilepsy. \u003cem\u003eClin. Neurophysiol.\u003c/em\u003e \u003cb\u003e122\u003c/b\u003e, 664\u0026ndash;671 (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSalami, P. et al. Dynamics of interictal spikes and high-frequency oscillations during epileptogenesis in temporal lobe epilepsy. \u003cem\u003eNeurobiol. Dis.\u003c/em\u003e \u003cb\u003e67\u003c/b\u003e, 97\u0026ndash;106 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShi, W. et al. Spike ripples localize the epileptogenic zone best: an international intracranial study. \u003cem\u003eBrain\u003c/em\u003e \u003cb\u003e147\u003c/b\u003e, 2496\u0026ndash;2506 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFrauscher, B. et al. High-frequency oscillations: The state of clinical research. \u003cem\u003eEpilepsia\u003c/em\u003e \u003cb\u003e58\u003c/b\u003e, 1316\u0026ndash;1329 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJacobs, J. et al. High-frequency electroencephalographic oscillations correlate with outcome of epilepsy surgery. \u003cem\u003eAnn. Neurol.\u003c/em\u003e \u003cb\u003e67\u003c/b\u003e, 209\u0026ndash;220 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFrauscher, B. et al. High-Frequency Oscillations in the Normal Human Brain. \u003cem\u003eAnn. Neurol.\u003c/em\u003e \u003cb\u003e84\u003c/b\u003e, 374\u0026ndash;385 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEngel, J. J., Bragin, A., Staba, R. \u0026amp; Mody, I. High-frequency oscillations: what is normal and what is not? \u003cem\u003eEpilepsia\u003c/em\u003e 50, 598\u0026ndash;604 (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZweiphenning, W. J. E. M. et al. Correcting for physiological ripples improves epileptic focus identification and outcome prediction. \u003cem\u003eEpilepsia\u003c/em\u003e \u003cb\u003e63\u003c/b\u003e, 483\u0026ndash;496 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBruder, J. C. et al. Physiological Ripples Associated with Sleep Spindles Differ in Waveform Morphology from Epileptic Ripples. \u003cem\u003eInt. J. Neural Syst.\u003c/em\u003e \u003cb\u003e27\u003c/b\u003e, 1750011 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShi, W. et al. Spike ripples localize the epileptogenic zone best: an international intracranial study. \u003cem\u003eBrain\u003c/em\u003e \u003cb\u003e147\u003c/b\u003e, 2496\u0026ndash;2506 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmiri, M., Lina, J. M., Pizzo, F. \u0026amp; Gotman, J. High Frequency Oscillations and spikes: Separating real HFOs from false oscillations. \u003cem\u003eClin. Neurophysiol.\u003c/em\u003e \u003cb\u003e127\u003c/b\u003e, 187\u0026ndash;196 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJacobs, J. et al. The identification of distinct high-frequency oscillations during spikes delineates the seizure onset zone better than high-frequency spectral power changes. \u003cem\u003eClin. Neurophysiol.\u003c/em\u003e \u003cb\u003e127\u003c/b\u003e, 129\u0026ndash;142 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlamoudi, O. A., Ilyas, A., Pati, S. \u0026amp; Iasemidis, L. Interictal localization of the epileptogenic zone: Utilizing the observed resonance behavior in the spectral band of surrounding inhibition. \u003cem\u003eFrontiers Neuroscience\u003c/em\u003e \u003cb\u003e16\u0026ndash;2022\u003c/b\u003e, (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHeers, M. et al. Spectral bandwidth of interictal fast epileptic activity characterizes the seizure onset zone. \u003cem\u003eNeuroImage: Clin.\u003c/em\u003e \u003cb\u003e17\u003c/b\u003e, 865\u0026ndash;872 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePartamian, H. et al. Machine learning on interictal intracranial EEG predicts surgical outcome in drug resistant epilepsy. \u003cem\u003enpj Digit. Med.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, 138 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorrell, G. A. et al. Recording and analysis techniques for high-frequency oscillations. \u003cem\u003eProg Neurobiol.\u003c/em\u003e \u003cb\u003e98\u003c/b\u003e, 265\u0026ndash;278 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMercier, M. R. et al. Advances in human intracranial electroencephalography research, guidelines and good practices. \u003cem\u003eNeuroImage\u003c/em\u003e \u003cb\u003e260\u003c/b\u003e, 119438 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMatarrese, M. A. G. et al. Spike propagation mapping reveals effective connectivity and predicts surgical outcome in epilepsy. \u003cem\u003eBrain\u003c/em\u003e \u003cb\u003e146\u003c/b\u003e, 3898\u0026ndash;3912 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDiamond, J. M., Chapeton, J. I., Theodore, W. H., Inati, S. K. \u0026amp; Zaghloul, K. A. The seizure onset zone drives state-dependent epileptiform activity in susceptible brain regions. \u003cem\u003eClin. Neurophysiol.\u003c/em\u003e \u003cb\u003e130\u003c/b\u003e, 1628\u0026ndash;1641 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTamilia, E. et al. Noninvasive Mapping of Ripple Onset Predicts Outcome in Epilepsy Surgery. \u003cem\u003eAnn. Neurol.\u003c/em\u003e \u003cb\u003e89\u003c/b\u003e, 911\u0026ndash;925 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlarcon, G. et al. Origin and propagation of interictal discharges in the acute electrocorticogram. Implications for pathophysiology and surgical treatment of temporal lobe epilepsy. \u003cem\u003eBrain\u003c/em\u003e \u003cb\u003e120\u003c/b\u003e, 2259\u0026ndash;2282 (1997).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLin, J. et al. High frequency oscillation network dynamics predict outcome in non-palliative epilepsy surgery. \u003cem\u003eBrain Commun.\u003c/em\u003e \u003cb\u003e6\u003c/b\u003e, fcae032 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eConrad, E. C. et al. Spatial distribution of interictal spikes fluctuates over time and localizes seizure onset. \u003cem\u003eBrain\u003c/em\u003e \u003cb\u003e143\u003c/b\u003e, 554\u0026ndash;569 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGliske, S. V. et al. Variability in the location of high frequency oscillations during prolonged intracranial EEG recordings. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cb\u003e9\u003c/b\u003e, 2155 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCimbalnik, J. et al. Multi-feature localization of epileptic foci from interictal, intracranial EEG. \u003cem\u003eClin. Neurophysiol.\u003c/em\u003e \u003cb\u003e130\u003c/b\u003e, 1945\u0026ndash;1953 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJaber, K. et al. A spatial perturbation framework to validate implantation of the epileptogenic zone. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e, 5253 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJiang, X., Cai, Z., Gonsisko, C., Worrell, G. A. \u0026amp; He, B. Mapping epileptogenic brain using a unified spatial\u0026ndash;temporal\u0026ndash;spectral source imaging framework. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e 122, e2510015122 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMirza, B. et al. Machine Learning and Integrative Analysis of Biomedical Big Data. \u003cem\u003eGenes (Basel)\u003c/em\u003e \u003cb\u003e10\u003c/b\u003e, (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZitnik, M. et al. Machine Learning for Integrating Data in Biology and Medicine: Principles, Practice, and Opportunities. \u003cem\u003eInf. Fusion\u003c/em\u003e. \u003cb\u003e50\u003c/b\u003e, 71\u0026ndash;91 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShamas, M., Yeh, H. J., Fried, I., Engel, J. J. \u0026amp; Staba, R. J. High-rate leading spikes in propagating spike sequences predict seizure outcome in surgical patients with temporal lobe epilepsy. \u003cem\u003eBrain Commun.\u003c/em\u003e \u003cb\u003e5\u003c/b\u003e, fcad289 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKlimes, P. et al. Interictal stereo-electroencephalography features below 45 Hz are sufficient for correct localization of the epileptogenic zone and postsurgical outcome prediction. \u003cem\u003eEpilepsia\u003c/em\u003e \u003cb\u003e65\u003c/b\u003e, 2935\u0026ndash;2945 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRaghu, A. et al. Machine Learning-Based localization of the epileptogenic zone using High-Frequency oscillations from SEEG: A Real-World approach. \u003cem\u003eJ. Clin. Neurosci.\u003c/em\u003e \u003cb\u003e135\u003c/b\u003e, 111177 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDaida, A. et al. AI-based localization of the epileptogenic zone using intracranial EEG. \u003cem\u003eEpilepsia Open.\u003c/em\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/epi4.70130\u003c/span\u003e\u003cspan address=\"10.1002/epi4.70130\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, Y. et al. Automatic Localization of Seizure Onset Zone Based on Multi-Epileptogenic Biomarkers Analysis of Single-Contact from Interictal SEEG. \u003cem\u003eBioengineering\u003c/em\u003e 9, (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVaratharajah, Y. et al. Integrating artificial intelligence with real-time intracranial EEG monitoring to automate interictal identification of seizure onset zones in focal epilepsy. \u003cem\u003eJ. Neural Eng.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e, 046035 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVaratharajah, Y. et al. Inter-ictal Seizure Onset Zone localization using unsupervised clustering and Bayesian Filtering. in. \u003cem\u003e8th International IEEE/EMBS Conference on Neural Engineering (NER)\u003c/em\u003e 533\u0026ndash;539 (2017). (2017). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1109/NER.2017.8008407\u003c/span\u003e\u003cspan address=\"10.1109/NER.2017.8008407\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIslam, M. R., Zhao, X., Miao, Y., Sugano, H. \u0026amp; Tanaka, T. Epileptic seizure focus detection from interictal electroencephalogram: a survey. \u003cem\u003eCogn. Neurodyn.\u003c/em\u003e \u003cb\u003e17\u003c/b\u003e, 1\u0026ndash;23 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCimbalnik, J. et al. Multi-feature localization of epileptic foci from interictal, intracranial EEG. \u003cem\u003eClin. Neurophysiol.\u003c/em\u003e \u003cb\u003e130\u003c/b\u003e, 1945\u0026ndash;1953 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSch\u0026ouml;nberger, J. et al. Distinction of Physiologic and Epileptic Ripples: An Electrical Stimulation Study. \u003cem\u003eBrain Sci\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e, (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJanca, R. et al. Detection of interictal epileptiform discharges using signal envelope distribution modelling: application to epileptic and non-epileptic intracranial recordings. \u003cem\u003eBrain Topogr\u003c/em\u003e. \u003cb\u003e28\u003c/b\u003e, 172\u0026ndash;183 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWeiss, S. A. et al. Fast ripples reflect increased excitability that primes epileptiform spikes. \u003cem\u003eBrain Commun.\u003c/em\u003e \u003cb\u003e5\u003c/b\u003e, fcad242 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCho, J. R. et al. Resection of individually identified high-rate high-frequency oscillations region is associated with favorable outcome in neocortical epilepsy. \u003cem\u003eEpilepsia\u003c/em\u003e \u003cb\u003e55\u003c/b\u003e, 1872\u0026ndash;1883 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJahromi, S. et al. Overlap of spike and ripple propagation onset predicts surgical outcome in epilepsy. \u003cem\u003eAnn. Clin. Transl Neurol.\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e, 2530\u0026ndash;2547 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAndrews, J. P. et al. Association of Seizure Spread With Surgical Failure in Epilepsy. \u003cem\u003eJAMA Neurol.\u003c/em\u003e \u003cb\u003e76\u003c/b\u003e, 462\u0026ndash;469 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAndrews, J. P. et al. Early seizure spread and epilepsy surgery: A systematic review. \u003cem\u003eEpilepsia\u003c/em\u003e \u003cb\u003e61\u003c/b\u003e, 2163\u0026ndash;2172 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, A. et al. Neural fragility as an EEG marker of the seizure onset zone. \u003cem\u003eNat. Neurosci.\u003c/em\u003e \u003cb\u003e24\u003c/b\u003e, 1465\u0026ndash;1474 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCorona, L. et al. Non-invasive mapping of epileptogenic networks predicts surgical outcome. \u003cem\u003eBrain\u003c/em\u003e \u003cb\u003e146\u003c/b\u003e, 1916\u0026ndash;1931 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMatarrese, M. A. G. et al. Spike propagation mapping reveals effective connectivity and predicts surgical outcome in epilepsy. \u003cem\u003eBrain\u003c/em\u003e \u003cb\u003e146\u003c/b\u003e, 3898\u0026ndash;3912 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGunnarsdottir, K. M. et al. Source-sink connectivity: a novel interictal EEG marker for seizure localization. \u003cem\u003eBrain\u003c/em\u003e \u003cb\u003e145\u003c/b\u003e, 3901\u0026ndash;3915 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIslam, M. R., Zhao, X., Miao, Y., Sugano, H. \u0026amp; Tanaka, T. Epileptic seizure focus detection from interictal electroencephalogram: a survey. \u003cem\u003eCogn. Neurodyn.\u003c/em\u003e \u003cb\u003e17\u003c/b\u003e, 1\u0026ndash;23 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMiao, Y., Iimura, Y., Sugano, H., Fukumori, K. \u0026amp; Tanaka, T. Seizure onset zone identification using phase-amplitude coupling and multiple machine learning approaches for interictal electrocorticogram. \u003cem\u003eCogn. Neurodyn.\u003c/em\u003e \u003cb\u003e17\u003c/b\u003e, 1591\u0026ndash;1607 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAkter, M. S. et al. Statistical Features in High-Frequency Bands of Interictal iEEG Work Efficiently in Identifying the Seizure Onset Zone in Patients with Focal Epilepsy. \u003cem\u003eEntropy\u003c/em\u003e 22, (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFabbri, L. et al. Noninvasive classification of physiological and pathological high frequency oscillations in children. \u003cem\u003eBrain Commun.\u003c/em\u003e \u003cb\u003e7\u003c/b\u003e, fcaf170 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMooij, A. H. et al. Accurate differentiation between physiological and pathological ripples recorded with scalp-EEG. \u003cem\u003eClin. Neurophysiol.\u003c/em\u003e \u003cb\u003e143\u003c/b\u003e, 172\u0026ndash;181 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKarpychev, V. et al. Epileptogenic high-frequency oscillations present larger amplitude both in mesial temporal and neocortical regions. \u003cem\u003eFront. Hum. Neurosci.\u003c/em\u003e \u003cb\u003e16\u003c/b\u003e, 984306 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCimbalnik, J. et al. Physiological and pathological high frequency oscillations in focal epilepsy. \u003cem\u003eAnn. Clin. Transl. Neurol.\u003c/em\u003e \u003cb\u003e5\u003c/b\u003e, 1062\u0026ndash;1076 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlkawadri, R. et al. The spatial and signal characteristics of physiologic high frequency oscillations. \u003cem\u003eEpilepsia\u003c/em\u003e \u003cb\u003e55\u003c/b\u003e, 1986\u0026ndash;1995 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen, C. et al. A region-specific modulation of sleep slow waves on interictal epilepsy markers in focal epilepsy. \u003cem\u003eEpilepsia\u003c/em\u003e \u003cb\u003e64\u003c/b\u003e, 973\u0026ndash;985 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLundstrom, B. N., Brinkmann, B. H. \u0026amp; Worrell, G. A. Low frequency novel interictal EEG biomarker for localizing seizures and predicting outcomes. \u003cem\u003eBrain Commun.\u003c/em\u003e \u003cb\u003e3\u003c/b\u003e, fcab231 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCorona, L. et al. Non-invasive mapping of epileptogenic networks predicts surgical outcome. \u003cem\u003eBrain\u003c/em\u003e \u003cb\u003e146\u003c/b\u003e, 1916\u0026ndash;1931 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShah, P. et al. High interictal connectivity within the resection zone is associated with favorable post-surgical outcomes in focal epilepsy patients. \u003cem\u003eNeuroimage Clin.\u003c/em\u003e \u003cb\u003e23\u003c/b\u003e, 101908 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGunnarsdottir, K. M. et al. Source-sink connectivity: a novel interictal EEG marker for seizure localization. \u003cem\u003eBrain\u003c/em\u003e \u003cb\u003e145\u003c/b\u003e, 3901\u0026ndash;3915 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRijal, S. et al. Functional connectivity discriminates epileptogenic states and predicts surgical outcome in children with drug resistant epilepsy. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cb\u003e13\u003c/b\u003e, 9622 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJacobs, J. et al. Removing high-frequency oscillations: A prospective multicenter study on seizure outcome. \u003cem\u003eNeurology\u003c/em\u003e \u003cb\u003e91\u003c/b\u003e, e1040\u0026ndash;e1052 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTamilia, E. et al. Surgical resection of ripple onset predicts outcome in pediatric epilepsy. \u003cem\u003eAnn. Neurol.\u003c/em\u003e \u003cb\u003e84\u003c/b\u003e, 331\u0026ndash;346 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNissen, I. A. et al. Optimization of epilepsy surgery through virtual resections on individual structural brain networks. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e, 19025 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEngel, J. Jr Surgery for seizures. \u003cem\u003eN. Engl. J. Med.\u003c/em\u003e \u003cb\u003e334\u003c/b\u003e, 647\u0026ndash;653 (1996).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDimakopoulos, V. et al. Protocol for multicentre comparison of interictal high-frequency oscillations as a predictor of seizure freedom. \u003cem\u003eBrain Commun.\u003c/em\u003e \u003cb\u003e4\u003c/b\u003e, fcac151 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFrauscher, B. et al. Facilitation of epileptic activity during sleep is mediated by high amplitude slow waves. \u003cem\u003eBrain\u003c/em\u003e \u003cb\u003e138\u003c/b\u003e, 1629\u0026ndash;1641 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSammaritano, M., Gigli, G. L. \u0026amp; Gotman, J. Interictal spiking during wakefulness and sleep and the localization of foci in temporal lobe epilepsy. \u003cem\u003eNeurology\u003c/em\u003e \u003cb\u003e41\u003c/b\u003e, 290\u0026ndash;290 (1991).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTadel, F., Baillet, S., Mosher, J. C., Pantazis, D. \u0026amp; Leahy, R. M. Brainstorm: a user-friendly application for MEG/EEG analysis. \u003cem\u003eComputational intelligence and neuroscience\u003c/em\u003e 1\u0026ndash;13 (2011). (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ede Rodrigues, M. et al. K. A FreeSurfer-Compliant Consistent Manual Segmentation of Infant Brains Spanning the 0\u0026ndash;2 Year Age Range. Front. Hum. Neurosci. 9, 21. (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoberts, D. W., Hartov, A., Kennedy, F. E., Miga, M. I. \u0026amp; Paulsen, K. D. Intraoperative brain shift and deformation: a quantitative analysis of cortical displacement in 28 cases. \u003cem\u003eNeurosurgery\u003c/em\u003e \u003cb\u003e43\u003c/b\u003e, 749\u0026ndash;758 (1998).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAzeem, A. et al. Interictal spike networks predict surgical outcome in patients with drug-resistant focal epilepsy. \u003cem\u003eAnn. Clin. Transl Neurol.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, 1212\u0026ndash;1223 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eConrad, E. C. et al. Spike patterns surrounding sleep and seizures localize the seizure-onset zone in focal epilepsy. \u003cem\u003eEpilepsia\u003c/em\u003e \u003cb\u003e64\u003c/b\u003e, 754\u0026ndash;768 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMaher, J. M., Markey, J. C. \u0026amp; Ebert-May, D. The Other Half of the Story: Effect Size Analysis in Quantitative Research. \u003cem\u003eCBE\u0026mdash;Life Sci. Educ.\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e, 345\u0026ndash;351 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOno, M., Kubik, S. \u0026amp; Abernathey, C. D. \u003cem\u003eAtlas of the Cerebral Sulci\u003c/em\u003e (G. Thieme, 1990).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"interictal epilepsy biomarkers, epilepsy surgery, combined interictal features, supervised machine learning","lastPublishedDoi":"10.21203/rs.3.rs-8682213/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8682213/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eDelineating the epileptogenic zone (EZ) is essential for achieving seizure freedom in drug-resistant epilepsy (DRE). Conventionally, seizure onset derived from ictal intracranial EEG (iEEG) approximates the EZ, but acquiring ictal data can be challenging. Interictal iEEG abnormalities offer abundant, easily acquired, non-seizure-dependent markers of the epileptogenic tissue; however, these biomarkers offer limited specificity. Here, we propose a machine-learning framework that integrates interictal spike and ripple features to automatically delineate the EZ and predict outcome with improved performance compared to individual biomarkers. We retrospectively analyzed iEEG data from 62 children with DRE ([34 good (Engel 1) outcomes] undergoing neurosurgery, automatically detected spikes and ripples, and computed temporal, spectral, and spatial features for each channel. We trained Random Forest classifiers to predict the EZ using combinations of these features. The predicted EZ derived from spike-based and combined spike-ripple feature sets outperformed those from individual biomarkers in defining the EZ, with an area under the receiver operating characteristic curve of 0.9 and 74% spatial overlap with resection. Although most individual features and classifiers predicted the outcome, the combined feature model performed best (i.e., sensitivity 88%, specificity 68%, and accuracy 79%). Our findings demonstrate that integrating multimodal interictal features improves the EZ delineation, providing valuable prognostic insights for epilepsy surgery.\u003c/p\u003e","manuscriptTitle":"Predicting Surgical Outcome in Drug-Resistant Epilepsy by Combining Interictal Biomarkers within a Machine Learning Framework","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-17 14:35:59","doi":"10.21203/rs.3.rs-8682213/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-02-25T07:51:25+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-22T05:59:52+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-21T05:28:12+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"80386671716470423070722271241844760150","date":"2026-02-14T04:38:04+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"156709522777500006843685750780450701905","date":"2026-02-13T03:39:41+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"148884303036645497529602162490656851128","date":"2026-02-12T07:57:12+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-02-12T04:32:32+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-02-11T17:30:41+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-02-03T12:49:19+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-01-29T17:56:32+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2026-01-29T17:43:44+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"3c70d962-1f42-46cf-8b0d-4a48ff4b571b","owner":[],"postedDate":"February 17th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":62917740,"name":"Health sciences/Biomarkers"},{"id":62917741,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":62917742,"name":"Health sciences/Neurology"},{"id":62917743,"name":"Biological sciences/Neuroscience"}],"tags":[],"updatedAt":"2026-03-30T16:22:03+00:00","versionOfRecord":{"articleIdentity":"rs-8682213","link":"https://doi.org/10.1038/s41598-026-45177-x","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2026-03-26 16:12:21","publishedOnDateReadable":"March 26th, 2026"},"versionCreatedAt":"2026-02-17 14:35:59","video":"","vorDoi":"10.1038/s41598-026-45177-x","vorDoiUrl":"https://doi.org/10.1038/s41598-026-45177-x","workflowStages":[]},"version":"v1","identity":"rs-8682213","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8682213","identity":"rs-8682213","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}