Enhancing Prediction of Individualized Antipsychotic Outcome with fMRI-EEG Feature Integration in First-Episode Schizophrenia: A Real-World Study

Enhancing Prediction of Individualized Antipsychotic Outcome with fMRI-EEG Feature Integration in First-Episode Schizophrenia: A Real-World Study | 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 Enhancing Prediction of Individualized Antipsychotic Outcome with fMRI-EEG Feature Integration in First-Episode Schizophrenia: A Real-World Study Xiang-Yang Zhang, Liju Liu, Dongmei Wang, Meng Chen, Xiaoe Lang, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6047108/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Predicting treatment efficacy in schizophrenia within real-world settings is of great clinical significance but challenging. fMRI and EEG reveal distinct spatial and temporal characteristics of brain information processing. Therefore, developing predictive models that integrate spatiotemporal multimodal features and accommodate the complexities of real-world environments is crucial for achieving accurate treatment outcome predictions and personalized therapy. Ninety first-episode, drug-naive schizophrenia patients underwent fMRI and EEG at baseline and received 6–8 weeks of single antipsychotic treatment in a naturalistic setting. Clinical symptoms were evaluated using the Positive and Negative Syndrome Scale (PANSS) at baseline and post-treatment. Entropy metrics reflecting information processing capacity were calculated from BOLD signals as fMRI features, while key ERP components (P50, N100, P200, N200, and P300) representing different cognitive stages were extracted with their amplitudes and latencies as EEG features. LASSO regression model was used to assess the predictive power of unimodal and multimodal features for PANSS score reduction. The multimodal model outperformed unimodal models in predicting improvements in PANSS total score (R = 0.440, P < 0.001), negative symptom (R = 0.328, P = 0.002), and general psychopathology (R = 0.449, P < 0.001). Key features in the multimodal model included WPE from the medial frontal gyrus, supplementary motor area, amygdala, caudate, pallidum, and thalamus, along with ERP characteristics like P50 ratio, N100, N200, P3a latencies, and P200 amplitude. These features were not significantly correlated, highlighting their complementary roles in information processing as key to the multimodal model's improved performance. This study demonstrates that multimodal fusion prediction model effectively integrates brain information processing features across different dimensions, significantly improving individualized prediction of treatment outcomes. The results hold important value for its translational application in precision psychiatry within real-world settings. Health sciences/Biomarkers/Prognostic markers Health sciences/Diseases/Psychiatric disorders/Schizophrenia Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Schizophrenia, as a severe mental disorder, presents with a wide range of clinical manifestations, including cognitive dysfunction, emotional instability, and social impairment[ 1 ]. Although antipsychotic medication treatment remains the primary treatment approach for schizophrenia, the response to medication varies significantly among patients[ 2 ], which complicates the development of personalized treatment plans. This challenge is particularly pronounced in real-world settings, where treatment outcomes are influenced by various factors, including individual physiological characteristics, the diversity of therapeutic approaches, and changes in the external environment. Therefore, accurately predicting the effects of personalized pharmacotherapy under conditions more reflective of clinical practice is crucial. In recent years, the rapid advancement of neuroimaging techniques has provided important tools for predicting treatment efficacy in schizophrenia. Among these, functional magnetic resonance imaging (fMRI) and electroencephalography (EEG), as complementary neuroimaging modalities, reveal different aspects of brain information processing from spatial and temporal dimensions, respectively, laying the foundation for exploring the neurobiological mechanisms underlying treatment prediction. fMRI, with its high spatial resolution, captures functional connectivity between brain regions, resting-state network properties, and spatial patterns of local brain activity. For example, Cao et al. and Sarpal et al. found that functional connectivity in the superior temporal cortex and the striatum could predict the treatment response to antipsychotic medications in first-episode drug-naive (FEDN) schizophrenia patients[ 3 , 4 ]. Blessing et al. reported that anterior hippocampal-cortical functional connectivity could distinguish patients from healthy controls and predict the response to second-generation antipsychotics[ 5 ]. Li et al. observed increased fractional amplitude of low frequency fluctuation in the left ventral striatum in FEDN schizophrenia patients across two independent samples, which predicted individual treatment response to olanzapine[ 6 ]. Additionally, a previous study of ours found reduced complexity in the functional activity of the striatum in FEDN schizophrenia patients, with baseline values significantly correlating with clinical symptom improvement after treatment with risperidone[ 7 ]. These studies indicate that fMRI can capture spatial pattern features of brain function, providing important clues for treatment prediction. EEG, with its high temporal resolution, captures dynamic changes in neurophysiological activity during brain information processing and has been successfully employed in several studies to predict patients' responses to clozapine treatment[ 8 – 10 ]. Event-related potentials (ERPs), an important component of EEG, serve as sensitive neurophysiological markers for real-time monitoring of brain responses to information processing. Among these, the reduction in P300 amplitude and the delay in latency are characteristic features of schizophrenia patients, which not only worsen with disease progression[ 11 ] but also track changes in clinical symptoms[ 12 , 13 ]. P50, a key indicator of sensory gating ability, reflects the brain's capacity to filter irrelevant information at the perceptual level. Nagamoto et al. found that P50 gating ability was closely related to the response to clozapine treatment[ 14 ], with atypical antipsychotics outperforming traditional antipsychotics in enhancing P50 gating[ 15 , 16 ]. These findings suggest that EEG holds promising potential for reflecting dynamic neural activity and treatment response in schizophrenia patients. Although both fMRI and EEG have shown certain advantages in unimodal efficacy prediction studies, single-modality data often fail to comprehensively reflect the complex neurobiological mechanisms of schizophrenia. In real-world settings, the diversity of treatment regimens (including medication types and dosages) further complicates prediction. Atypical antipsychotic drugs, due to their lower side effects and broader applicability[ 17 – 19 ], particularly their significant effects on improving cognitive and emotional symptoms[ 20 – 23 ], have become first-line treatment options. However, no study has yet integrated fMRI and EEG features to predict the individualized treatment response to atypical antipsychotics in FEDN schizophrenia patients within real-world settings. Therefore, there is an urgent need to explore whether the combination of these two modalities can significantly enhance the accuracy of efficacy prediction. In this study, we investigate whether combined fMRI and EEG features can predict individualized clinical outcomes in FEDN schizophrenia patients in real-world settings and assess the potential improvement in predictive accuracy compared to unimodal features. The study incorporates personalized treatment regimens without restrictions on medication types or dosages, thus more comprehensively reflecting the complexity of real-world treatment and enhancing the clinical applicability of the model. By integrating multimodal features across both spatial and temporal dimensions, this research provides a theoretical basis for uncovering the neurobiological mechanisms underlying treatment response in schizophrenia patients and for advancing the clinical translation of multimodal imaging technologies. Methods Participants This study included 90 FEDN schizophrenia outpatients recruited from the First Hospital of Shanxi Medical University. The inclusion criteria were as follows: 1) met the DSM-IV diagnostic criteria for schizophrenia, with the diagnosis established by two trained psychiatrists based on structured clinical interview for DSM-IV (SCID)[ 24 ]; 2) Positive and Negative Syndrome Scale (PANSS) score[ 25 ] ≥ 60; 3) no prior treatment with psychotropic medications; 4) age < 45 years, Han nationality; and 5) illness duration no longer than 5 years. The exclusion criteria were: 1) presence of metabolic diseases such as diabetes and hyperlipidemia, organic brain diseases, or a history of traumatic brain injury; 2) hearing or visual impairments that could affect brain evoked potential examinations; and 3) other psychiatric disorders or substance dependence (except for nicotine) according to DSM-IV criteria. All participants provided written informed consent, and the study protocol was approved by the Institutional Review Board of the First Hospital of Shanxi Medical University. Medication and Clinical Assessments All patients received second-generation antipsychotic medications (e.g., olanzapine, risperidone, or aripiprazole) as prescribed by their attending physicians based on individual clinical conditions, without the inclusion of psychotherapy or other non-pharmacological interventions. The selection of treatment regimens and dosage adjustments were made according to the patient's condition, drug tolerance, and the clinical judgment of the physician. All patients underwent a second clinical assessment between the 6th and 8th week after the initiation of treatment. Clinical symptoms were assessed using the 30-item PANSS. fMRI Data Acquisition and Preprocessing Resting-state fMRI data were acquired on a General Electric Signa HDxt 3.0 T scanner with a 32-channel head coil at baseline. Functional data were obtained using a gradient-echo echo-planar imaging sequence [TR/TE = 2000ms/30ms, flip angle = 90°, field of view = 240×240 mm 2 , matrix size = 64 × 64; slice thickness = 4mm, voxel size = 3.75 × 3.75 × 4 mm 3 , 210 time points collected]. The fMRI data preprocessing was performed using the Data Processing Assistant for Resting-State fMRI (DPARSF, V5.4, http://rfmri.org/DPARSF ). To minimize magnetization equilibrium effects, the first 10 volumes were discarded. Slice timing correction and realignment were applied to the remaining images. The functional images were then normalized to the EPI template with a resampled voxel size of 3mm × 3mm × 3mm. Several covariates, including Friston 24 motion parameters, cerebrospinal fluid, and white matter signals, were regressed out as nuisance variables to reduce spurious variance. Spatial smoothing (FWHM = 6mm), detrending, and band-pass filtering (0.01–0.08 Hz) were applied to the data. Participants were excluded from further analysis if their translational or rotational displacement exceeded 3.0 mm or 3.0°, or if their mean frame-wise displacement (a measure of micromovement) exceeded 0.3 mm. Twelve participants were excluded due to excessive head motion, leaving 78 participants for the final feature extraction and prediction tasks. The demographic and clinical characteristics of participants included for analysis are shown in supplementary Table S1 . Calculation of WPE Features Weighted Permutation Entropy (WPE) is a measure of time series complexity that quantifies the unpredictability and irregularity of the signal. It calculates the degree of randomness in the data by examining the order of values within the time series. Higher WPE values indicate more complex and unpredictable dynamics, while lower values suggest more regular patterns. In this study, regional BOLD time series were extracted using the Anatomical Automatic Labeling (AAL) atlas[ 26 ], which defines 90 distinct cerebral functional regions, excluding the cerebellum. The WPE for each brain region was then calculated based on its entire time series. Detailed information on the WPE calculation method can be found in our previous studies[ 7 ]. Paradigm The P50 components were measured using a conditioning-testing paradigm. Participants were instructed to relax and focus on a fixed point. We generated a stimulus signal of 90-dB pulses of 0.1ms in duration and recorded the event-related potential waveforms. A total of 32 pairs of auditory stimuli were presented, with an inter-stimulus interval of 500ms and an inter-pair interval of 10s[ 27 , 28 ]. Each epoch lasted for 1000ms, including a 100ms before stimulation 1, 500ms between stimuli and 400ms after stimulation 2. The N100, P200, N200, and P300 components were measured using an auditory oddball paradigm. Participants were presented with frequent standard tones (500 Hz, 80%), infrequent target tones (1000 Hz, 10%), and infrequent novel distractor sounds (variety of sounds, 10%). Standard and target tones were 50 milliseconds in duration (5-millisecond rise/fall time) at an 80-dB sound pressure level. Stimuli were presented with a 1.25-second stimulus onset asynchrony. Participants were instructed to press a response button with their preferred hand upon hearing the target stimuli, while ignoring the standard and novel sounds. EEG Data Acquisition and Preprocessing A signal generator and a digital 40-channel EBNeuro Sirius EEG system (EBNeuro, Florence, Italy) were used for electrophysiological recordings and EEG signal processing. EEG signals were filtered using a 0.1–250 Hz analog filter and a 50 Hz notch filter and sampled at a frequency of 500 Hz. Eye movements were recorded using Ag-AgCl disc electrodes to capture electro-oculography (EOG) signals, and ocular artifacts were corrected offline using the independent component analysis method. Vertical EOG signals were recorded with electrodes placed above and below the left eye, while horizontal EOG signals were recorded with electrodes placed at the outer canthi of both eyes. The ground electrode was positioned at the midpoint between FPz and Fz, and the reference electrode was placed on the right mastoid. Offline correction of reference voltage was conducted using the algebraic average of the left and right mastoids. The impedance of all electrodes was maintained below 5kΩ. Trials with artifacts exceeding ± 50µV were excluded. Offline analysis was conducted using bandpass filtering (10–50 Hz, 12 dB/octave roll-off). To control background noise during the experiment, continuous 70 dB(A) broadband white noise was provided throughout the recording. For the P50 component, S1 and S2 amplitudes and latencies were extracted from the Cz electrode. The S1 amplitude was defined as the maximum peak within the 40–80ms time window following the S1 stimulus onset, while the S2 amplitude was determined at the latency window closest to the S1 latency after the S2 stimulus. The P50 ratio, an index of sensory gating, was calculated by dividing the S2 amplitude by the S1 amplitude. For the N100, P200, N200, and P300 components, individual peak amplitudes and latencies were assessed within the following time windows: N100 (60-180ms), P200 (150-250ms), N200 (200-350ms), and P300 (230-420ms). The amplitudes and latencies of N100, P200, and N200 components were extracted from the Cz electrode. The P3b peak amplitudes and latencies were chosen from the target minus standard difference wave as the most positive peak at electrode Pz (where P3b is maximum), while P3a peak amplitudes and latencies were identified from the novel minus standard difference wave at electrode Cz (where P3a is maximum). All ERP components were extracted by two independent neurophysiologists who were blinded to the participants' status. In cases of discrepancies in the analysis results, a third party conducted further evaluations to ensure consistency and accuracy. Prediction of Treatment Outcome with WPE and EEG Measures This study aims to investigate whether EEG features and WPE features can predict changes in PANSS scores for patients with schizophrenia under real-world settings. Using the LASSO regression model combined with cross-validation, EEG features (15), WPE features (90 regions), and their combination were used as predictors to assess their ability to predict changes in the PANSS total score, positive, negative, and general symptom subscales (PANSS scores at baseline minus PANSS scores at follow-up). We also compared the predictive performance of single-modal (EEG or WPE) and multi-modal features. All features were standardized before entering the model to eliminate scale differences and ensure comparisons on the same scale. Additionally, to control for potential confounding factors, feature values were adjusted for covariates, including patient sex, age, and illness duration, prior to standardization. LASSO regression is an L1-norm regularization method that introduces a shrinkage penalty term (λ) to avoid overfitting and automatically shrinks the coefficients of less important features to zero. It is particularly suitable for datasets with high multicollinearity. Model evaluation was performed using a repeated nested cross-validation method (10 outer folds and 10 inner folds). In each outer fold, the inner folds were used to optimize λ, and predictions were made on the test set. After cross-validation, the predicted reduction scores for each subject were compared to the actual reduction scores. Model performance was evaluated by the correlation between predicted and observed values, and significance was determined using 1000 permutations. In each permutation, subject labels were randomly shuffled, and the entire cross-validation procedure was repeated on the shuffled data. The P-value was calculated as the proportion of correlation coefficients in the null models greater than the observed ones. Results Prediction of PANSS-T Score As shown in Figure. 1A, using WPE features as predictors, the cross-validated LASSO regression revealed significant correlations between predicted and actual reductions in the PANSS-T score ( R = 0.288, P = 0.003). The selected features for the model are listed on the right. Specifically, the top features contributing to symptom changes (features selected across all 10 outer folds) included the bilateral orbital part of middle frontal gyrus, right triangular part of inferior frontal gyrus, left supplementary motor area, left medial orbital part of superior frontal gyrus, bilateral amygdala, left calcarine, right lingual, right inferior occipital gyrus, right superior parietal gurus, bilateral supramarginal gyrus, left precuneus, bilateral caudate, right thalamus and right temporal pole. As shown in Figure. 1B, using EEG features as predictors, the cross-validated LASSO regression revealed significant correlations between predicted and actual reductions in the PANSS-T score ( R = 0.401, P = 0.002). Note that the result remained highly significant ( R = 0.344, P = 0.001) even after removing an outlier at the top right of the scatter plot, as shown in supplementary Figure. S1A. The top features contributing to symptom changes included the amplitude of P200, latency of N200, amplitude of P3b and P50 ratio. As shown in Fig. 1 C, the LASSO regression model using the combination of WPE and EEG features predicted the PANSS total score with a significant correlation ( R = 0.440, P < 0.001). The top features contributing to symptom changes included the left middle frontal gyrus, bilateral orbital part of middle frontal gyrus, right triangular part of inferior frontal gyrus, right amygdala, left calcarine, right inferior occipital gyrus, right supramarginal gyrus, bilateral caudate, right thalamus, left Heschl gyrus, latency of N100, amplitude of P200, latency of N200 and P50 ratio. Prediction of PANSS-N Score As shown in Figure. 2A, using WPE features as predictors, the cross-validated LASSO regression revealed significant correlations between predicted and actual reductions in the PANSS-N score ( R = 0.232, P = 0.021). The top features contributing to symptom changes included the left anterior cingulate and paracingulate gyrus, left middle occipital gyrus, right inferior occipital gyrus, right supramarginal gyrus, bilateral caudate and left pallidum. As shown in Figure. 2B, using EEG features as predictors, the cross-validated LASSO regression revealed significant correlations between predicted and actual reductions in the PANSS-N score ( R = 0.288, P = 0.008). The top features contributing to symptom changes included the amplitude of P200, latency of N200 and latency of P3a. However, the result was no longer significant after excluding the outlier in the top-right corner. As shown in Fig. 2 C, the LASSO regression model using the combination of WPE and EEG features predicted the PANSS-N score with a significant correlation ( R = 0.328, P = 0.002). The top features contributing to symptom changes included the left middle occipital gyrus, right supramarginal gyrus, left caudate left pallidum, amplitude of P200 and latency of P3a. Prediction of PANSS-P Score Neither single-modality features nor the combined features effectively predicted the improvement in positive symptoms after treatment. Using WPE features, the prediction result was R = 0.129, P = 0.148; with EEG features, the result was R = 0.121, P = 0.147; and with combined features, the prediction result was R = 0.165, P = 0.074, which is close to statistical significance but does not reach a significant level. Prediction of PANSS-G Score As shown in Figure. 3A, using WPE features as predictors, the cross-validated LASSO regression revealed significant correlations between predicted and actual reductions in the PANSS-G score ( R = 0.293, P = 0.002). Note that the result remained highly significant ( R = 0.267, P < 0.001) even after removing an outlier at the top right of the scatter plot, as shown in supplementary Figure. S1B. The top features contributing to symptom changes include the middle frontal gyrus, left orbital part of middle frontal gyrus, left supplementary motor area, right amygdala, right inferior occipital gyrus, right superior parietal gyrus, right inferior parietal, left caudate and right temporal pole. As shown in Figure. 3B, using EEG features as predictors, the cross-validated LASSO regression revealed significant correlations between predicted and actual reductions in the PANSS-N score ( R = 0.393, P = 0.001). Note that the result remained highly significant ( R = 0.378, P < 0.001) even after removing an outlier at the top right of the scatter plot, as shown in supplementary Figure. S1C. The top features contributing to symptom changes included the amplitude of P200, latency of N200, and P50 ratio. As shown in Fig. 3 C, the LASSO regression model using the combination of WPE and EEG features predicted the PANSS-N score with a significant correlation ( R = 0.449, P < 0.001). The top features contributing to symptom changes included the left Middle frontal gyrus, left orbital part of middle frontal gyrus, left supplementary motor area, right amygdala, left cuneus, right inferior occipital gyrus, right superior parietal gyrus, left caudate, latency of N200, and P50 ratio. Correlation between WPE and EEG Features In this study, we further investigated the correlation between WPE features and EEG features. Specifically, we extracted the top features selected in the LASSO model (features selected across all 10 outer folds) and computed the Pearson correlation between these WPE features and EEG features. The results showed that, except for the significant correlation between the WPE feature of the right amygdala and the P50 ratio ( R = 0.290, P uncorrected =0.010), no significant correlations were found between the other features, as shown in Fig. 4 . This result suggests that the improvement in the performance of the fusion model is not due to the correlation or redundancy between the two modal features, but rather because they provide complementary information that reflects different aspects of brain information processing. Discussion This study found that baseline fMRI and EEG features can effectively predict the improvement of psychiatric symptoms, including negative symptoms, general symptoms, and total symptom scores, in FEDN schizophrenia patients after 6–8 weeks of treatment with atypical antipsychotics. The multimodal fusion model demonstrated higher predictive capability compared to unimodal features, as evidenced by an improvement in the model's R-value and the lack of significant correlations between key features selected in the fusion model. This suggests that these features provide complementary information in reflecting the brain's information processing mechanisms. The selected key features include WPE characteristics from brain regions such as the middle frontal gyrus, triangular part of the inferior frontal gyrus, supplementary motor area, amygdala, caudate, pallidum, and thalamus, as well as latency of N100, N200, P3a, amplitude of P200, and P50 ratio, further revealing the potential roles of these features in predicting treatment outcomes. This study has several notable advantages: First, the participants included were all DNFE schizophrenia patients, ensuring that the fMRI and EEG features could more purely reflect the pathophysiological characteristics of schizophrenia itself, without the confounding effects of disease progression or previous pharmacological treatments. Second, this study was conducted in a real-world setting, where treatment regimens were not restricted by medication types or dosages, encompassing a variety of atypical antipsychotic treatments. This design more closely mirrors clinical practice, enhancing the model's generalizability and clinical applicability. Third, unlike previous studies that have focused on group-level binary classification of antipsychotic treatment response, this study emphasizes individualized efficacy prediction. Group-level responder/non-responder classification methods have certain limitations, as they rely on arbitrary classification criteria and overlook intra-group variability in treatment effects, which may complicate individualized treatment decisions for clinicians. In contrast, this study overcomes these limitations by predicting treatment efficacy as a continuous variable, providing a model with greater clinical reference value. Fourth, we are the first to validate the superiority and complementarity of combined EEG and fMRI features in efficacy prediction. By combining the high temporal resolution of EEG with the high spatial resolution of fMRI, this study revealed the synergistic effect of the two modalities across different dimensions of information processing, further enhancing the predictive power of treatment responses in schizophrenia patients and providing valuable support for the application of multimodal imaging technologies in schizophrenia research. Our study found that the complexity of brain regions, including the middle frontal gyrus, triangular part of the inferior frontal gyrus, supplementary motor area, amygdala, caudate, pallidum, and thalamus, significantly contributes to predicting clinical symptom changes after antipsychotic treatment. In schizophrenia-related research, brain activity complexity, as an indicator of information processing capacity[ 29 , 30 ], has provided unique insights into the neurobiological mechanisms of the disorder. The middle frontal gyrus and triangular part of the inferior frontal gyrus are critical components of the prefrontal cortex, and dysfunction in these areas is closely associated with symptoms such as impaired executive function, working memory deficits[ 31 – 33 ], and disorganization[ 34 ]. Numerous studies have reported abnormal complexity in the frontal lobe of schizophrenia patients, including increased BOLD signal complexity in the left inferior frontal gyrus and Brodmann area 47[ 35 ], as well as decreased complexity in the middle and superior frontal regions[ 36 , 37 ]. These findings collectively suggest a disruption in the information processing abilities of the prefrontal cortex. Research has shown that schizophrenia patients exhibit reduced amygdala volume[ 38 , 39 ], which correlates with lower cognitive function scores; mouse models indicate that activation of the prefrontal cortex-basolateral amygdala pathway can improve cognitive performance[ 40 ], suggesting that amygdala dysfunction plays a significant role in cognitive deficits in schizophrenia. The caudate, pallidum, and thalamus form a crucial part of the cortico-striato-pallido-thalamo-cortical circuit[ 41 , 42 ], which plays a central role in regulating cognitive functions and behavioral responses. Abnormalities in this circuit are thought to underlie the symptoms and cognitive deficits of schizophrenia[ 43 , 44 ]. In a previous study, we found that the complexity of brain activity in the caudate nucleus, putamen, and pallidum was significantly reduced in first-episode schizophrenia patients, and the complexity of the caudate nucleus was closely related to cognitive deficits such as attention maintenance and verbal fluency[ 7 ]. On the other hand, the thalamus, as a hub for information transmission and sensory integration, has low connectivity with the prefrontal cortex, which is associated with impairments in higher cognitive functions such as attention and working memory[ 45 , 46 ]. Xue et al. reported a significant reduction in the complexity of the bilateral thalamus in schizophrenia patients[ 47 ], suggesting that the dynamic information processing capacity of the thalamus may be impaired, which could affect the filtering and integration of multisensory information, further exacerbating cognitive deficits. In summary, the predictive model in this study successfully captured the key brain regions associated with functional abnormalities in schizophrenia, providing important biological evidence for individualized efficacy prediction. Our study also found that the latency of N100, N200, and P3a, the amplitude of P200, and the P50 ratio all significantly contributed to predicting clinical symptom changes following antipsychotic treatment. These ERP components reflect the neurophysiological activity at different stages of information processing in the brain, revealing multi-level impairments in perception, attention, and cognitive functions in schizophrenia patients. P50 gating deficits are a hallmark feature of schizophrenia[ 48 – 50 ], reflecting an impairment in sensory information filtering. In a previous study, we found a significant correlation between P50 gating and general psychopathology as well as PANSS total scores in first-episode schizophrenia patients[ 51 ], further supporting its value in predicting treatment response. Next, N100, an early response to sensory input, with reduced amplitude and delayed latency, indicates dysfunction in the auditory cortex and early deficits in selective attention in schizophrenia patients[ 52 , 53 ]. A meta-analysis showed that most studies reported reduced P200 amplitude in schizophrenia patients, while a few showed increased or no difference[ 54 ]. Changes in P200 amplitude further highlight potential abnormalities in stimulus encoding and selective attention in patients. These attention-related abnormalities are not limited to the early stages of information processing but may extend to subsequent sensory gating processes related to attention[ 55 ], thereby affecting higher-order information processing efficiency. At later stages of cognitive processing, abnormalities in N200 and P300 are more prominent. N200 is primarily evoked by the frontal and central brain regions and is closely related to cognitive control, novelty processing, and matching processes[ 56 ]. Reduced amplitude and delayed latency of N200 suggest impairments in conflict monitoring and cognitive flexibility in patients[ 57 , 58 ]. P3a mainly reflects the attentional shift toward task-irrelevant stimuli or information[ 59 ]. Several studies have found that the auditory P3a response elicited in passive auditory oddball tasks is attenuated in psychotic-spectrum disorders and those at risk for psychosis[ 60 – 62 ]. These results suggest that the predictive model in our study successfully captured the multi-level abnormalities in perception, attention, and cognitive processing reflected by ERP components, validating their predictive value as neurophysiological biomarkers and providing support for the development of individualized treatment plans. It is noteworthy that our study found no significant correlation between the fMRI and EEG features extracted by the fusion model, suggesting that the independence and complementarity of these two modalities in reflecting different dimensions of information are key to enhancing the model's predictive performance. Several EEG-fMRI joint studies have shown that brain regions such as the thalamus, middle frontal gyrus, and dorsolateral prefrontal cortex are involved in P50 gating[ 63 – 65 ]. Tregellas et al. found that P50 gating ability was significantly correlated with BOLD signals in the thalamus[ 65 ], highlighting its important role in sensory information filtering. P300, an important ERP component associated with target detection and attentional shifting, is considered a marker of processing speed and efficiency, with its amplitude and latency reflecting these aspects. Horovitz et al. showed that in oddball tasks, the probability of target events not only influenced P300 amplitude and latency but also synchronously affected BOLD signal changes in related brain regions, such as the supramarginal gyri, thalamus, insula and right medial frontal gyrus. Additionally, Mulert et al. found consistent activations of fMRI and EEG signals in most regions during target detection tasks, including the temporo-parietal junction, supplementary motor area/anterior cingulate cortex, insula, and middle frontal gyrus[ 66 ]. These results emphasize the synergistic role of EEG and fMRI in cognitive processes. The advantages of the fusion model lie not only in the complementarity of temporal and spatial resolution but also in capturing the synergistic effects of static and dynamic information processing. EEG features reflect information filtering and target detection on a short time scale, while WPE indices reveal the complexity of functional brain activity, particularly in resource regulation and dynamic adaptation. This complementarity enables the fusion model to more comprehensively characterize information processing abnormalities in schizophrenia patients and combine the neurobiological mechanisms of temporal and spatial dimensions, providing important support for the efficiency and accuracy of the predictive model. The present study has some limitations. First, First, although the inclusion of DNFE schizophrenia patients minimizes the confounding effects of prior treatment, the relatively small sample size, particularly for a machine learning study focusing on individual prediction, may limit the generalizability of our findings to broader clinical populations. Second, although this study focused on treatment prediction in real-world settings, the lack of detailed information on the specific types and doses of antipsychotic medications prevented adjustment for these potential confounding factors before incorporating features into the model, which may impact the accuracy and interpretability of the predictions. Third, this study lacked an independent dataset to validate the performance improvements of the fusion model, which may limit its generalizability across different samples and clinical settings. In conclusion, this study provides initial evidence that the integration of fMRI-based complexity measures and EEG-derived ERP components enhances the prediction of antipsychotic treatment outcomes in DNFE schizophrenia patients within real-world settings. These findings underscore the advantage of multi-modal neuroimaging in capturing complementary information across spatial and temporal dimensions, offering valuable insights for advancing precision psychiatry and individualized treatment strategies. Declarations Supplementary Information Supplementary information is available at MP's website. Data Availability Data are available from the corresponding author upon reasonable request. Acknowledgements This research was funded by National Natural Science Foundation of China (62373079), Science and Technology Department of Sichuan Province (2024ZYD0039), Health Commission of Sichuan Province(24CXTD11), Sichuan Medical Association (S23012), Chengdu Science and Technology Bureau (2022-YF05-01867-SN), Health Commission of Chengdu (2024141), CAS International Cooperation Research Program (153111KYSB20190004) and STI2030-Major Projects 2021ZD0202102. C ompeting Interest s The authors declare no competing interests. CRediT A uthorship C ontribution S tatement Liju Liu: Methodology, Conceptualization, Visualization, Writing-original draft . Dongmei Wang : Supervision, Writing—review & editing. Meng Chen: Formal analysis, Writing-original draft. XiaoE Lang : Investigation, Writing—review & editing. Mi Yang: Resources, Writing-original draft, Writing-review & editing. Xiangyang Zhang: Investigation, Resources, Writing-review & editing. References Mueser KT, McGurk SR. Schizophrenia. The Lancet. 2004;363:2063–2072. Brenner HD, Dencker SJ, Goldstein MJ, Hubbard JW, Keegan DL, Kruger G, et al. Defining treatment refractoriness in schizophrenia. Schizophr Bull. 1990;16:551–561. Cao B, Cho RY, Chen D, Xiu M, Wang L, Soares JC, et al. Treatment response prediction and individualized identification of first-episode drug-naïve schizophrenia using brain functional connectivity. Mol Psychiatry. 2020;25:906–913. Sarpal DK, Argyelan M, Robinson DG, Szeszko PR, Karlsgodt KH, John M, et al. Baseline Striatal Functional Connectivity as a Predictor of Response to Antipsychotic Drug Treatment. Am J Psychiatry. 2016;173:69–77. Blessing EM, Murty VP, Zeng B, Wang J, Davachi L, Goff DC. Anterior Hippocampal-Cortical Functional Connectivity Distinguishes Antipsychotic Naïve First-Episode Psychosis Patients From Controls and May Predict Response to Second-Generation Antipsychotic Treatment. Schizophr Bull. 2020;46:680–689. Li H, Guo W, Liu F, Chen J, Su Q, Zhang Z, et al. Enhanced baseline activity in the left ventromedial putamen predicts individual treatment response in drug-naive, first-episode schizophrenia: Results from two independent study samples. EBioMedicine. 2019;46:248–255. Liu L, Li Z, Kong D, Huang Y, Wu D, Zhao H, et al. Neuroimaging markers of aberrant brain activity and treatment response in schizophrenia patients based on brain complexity. Transl Psychiatry. 2024;14:1–12. Khodayari-Rostamabad A, Hasey GM, Maccrimmon DJ, Reilly JP, de Bruin H. A pilot study to determine whether machine learning methodologies using pre-treatment electroencephalography can predict the symptomatic response to clozapine therapy. Clin Neurophysiol Off J Int Fed Clin Neurophysiol. 2010;121:1998–2006. Ravan M, Hasey G, Reilly JP, MacCrimmon D, Khodayari-Rostamabad A. A machine learning approach using auditory odd-ball responses to investigate the effect of Clozapine therapy. Clin Neurophysiol Off J Int Fed Clin Neurophysiol. 2015;126:721–730. Masychev K, Ciprian C, Ravan M, Manimaran A, Deshmukh A. Quantitative biomarkers to predict response to clozapine treatment using resting EEG data. Schizophr Res. 2020;223:289–296. Mathalon DH, Ford JM, Rosenbloom M, Pfefferbaum A. P300 reduction and prolongation with illness duration in schizophrenia. Biol Psychiatry. 2000;47:413–427. Ford JM, Mathalon DH, Marsh L, Faustman WO, Harris D, Hoff AL, et al. P300 amplitude is related to clinical state in severely and moderately ill patients with schizophrenia. Biol Psychiatry. 1999;46:94–101. Mathalon DH, Ford JM, Pfefferbaum A. Trait and state aspects of p300 amplitude reduction in schizophrenia: a retrospective longitudinal study. Biol Psychiatry. 2000;47:434–449. Nagamoto HT, Adler LE, McRae KA, Huettl P, Cawthra E, Gerhardt G, et al. Auditory P50 in schizophrenics on clozapine: improved gating parallels clinical improvement and changes in plasma 3-methoxy-4-hydroxyphenylglycol. Neuropsychobiology. 1999;39:10–17. Light GA, Geyer MA, Clementz BA, Cadenhead KS, Braff DL. Normal P50 Suppression in Schizophrenia Patients Treated With Atypical Antipsychotic Medications. Am J Psychiatry. 2000;157:767–771. Becker J, Gomes I, Ghisolfi ES, Schuch A, Ramos FLP, Ehlers JA, et al. Clozapine, but not typical antipsychotics, correct P50 suppression deficit in patients with schizophrenia. Clin Neurophysiol Off J Int Fed Clin Neurophysiol. 2004;115:396–401. Leucht S, Corves C, Arbter D, Engel RR, Li C, Davis JM. Second-generation versus first-generation antipsychotic drugs for schizophrenia: a meta-analysis. Lancet Lond Engl. 2009;373:31–41. Fabrazzo M, Cipolla S, Camerlengo A, Perris F, Catapano F. Second-Generation Antipsychotics’ Effectiveness and Tolerability: A Review of Real-World Studies in Patients with Schizophrenia and Related Disorders. J Clin Med. 2022;11:4530. Leucht S, Priller J, Davis JM. Antipsychotic Drugs: A Concise Review of History, Classification, Indications, Mechanism, Efficacy, Side Effects, Dosing, and Clinical Application. Am J Psychiatry. 2024;181:865–878. Bilder RM, Goldman RS, Volavka J, Czobor P, Hoptman M, Sheitman B, et al. Neurocognitive effects of clozapine, olanzapine, risperidone, and haloperidol in patients with chronic schizophrenia or schizoaffective disorder. Am J Psychiatry. 2002;159:1018–1028. Good KP, Kiss I, Buiteman C, Woodley H, Rui Q, Whitehorn D, et al. Improvement in cognitive functioning in patients with first-episode psychosis during treatment with quetiapine: an interim analysis. Br J Psychiatry Suppl. 2002;43:s45-49. Veselinović T, Scharpenberg M, Heinze M, Cordes J, Mühlbauer B, Juckel G, et al. Disparate effects of first and second generation antipsychotics on cognition in schizophrenia - Findings from the randomized NeSSy trial. Eur Neuropsychopharmacol J Eur Coll Neuropsychopharmacol. 2019;29:720–739. Torrisi SA, Laudani S, Contarini G, De Luca A, Geraci F, Managò F, et al. Dopamine, Cognitive Impairments and Second-Generation Antipsychotics: From Mechanistic Advances to More Personalized Treatments. Pharm Basel Switz. 2020;13:365. First M, Spitzer R, Gibbon M, Williams J. Structured clinical interview for DSM-IV-TR Axis I Disorders, Research Version, Non-patient Edition. SCID-IP, 2002. Kay SR, Fiszbein A, Opler LA. The positive and negative syndrome scale (PANSS) for schizophrenia. Schizophr Bull. 1987;13:261–276. Tzourio-Mazoyer N, Landeau B, Papathanassiou D, Crivello F, Etard O, Delcroix N, et al. Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. NeuroImage. 2002;15:273–289. Zhang XY, Liu L, Liu S, Hong X, Chen DC, Xiu MH, et al. Short-Term Tropisetron Treatment and Cognitive and P50 Auditory Gating Deficits in Schizophrenia. Am J Psychiatry. 2012;169:974–981. Xia L, Wang D, Wang J, Xu H, Huo L, Tian Y, et al. Association of cognitive and P50 suppression deficits in chronic patients with schizophrenia. Clin Neurophysiol Off J Int Fed Clin Neurophysiol. 2020;131:725–733. McDonough IM, Nashiro K. Network complexity as a measure of information processing across resting-state networks: evidence from the Human Connectome Project. Front Hum Neurosci. 2014;8. Wang DJJ, Jann K, Fan C, Qiao Y, Zang Y-F, Lu H, et al. Neurophysiological Basis of Multi-Scale Entropy of Brain Complexity and Its Relationship With Functional Connectivity. Front Neurosci. 2018;12. Perlstein WM, Carter CS, Noll DC, Cohen JD. Relation of prefrontal cortex dysfunction to working memory and symptoms in schizophrenia. Am J Psychiatry. 2001;158:1105–1113. Perlstein WM, Dixit NK, Carter CS, Noll DC, Cohen JD. Prefrontal cortex dysfunction mediates deficits in working memory and prepotent responding in schizophrenia. Biol Psychiatry. 2003;53:25–38. Van Snellenberg JX, Girgis RR, Horga G, van de Giessen E, Slifstein M, Ojeil N, et al. Mechanisms of Working Memory Impairment in Schizophrenia. Biol Psychiatry. 2016;80:617–626. MacDonald AW, Carter CS, Kerns JG, Ursu S, Barch DM, Holmes AJ, et al. Specificity of prefrontal dysfunction and context processing deficits to schizophrenia in never-medicated patients with first-episode psychosis. Am J Psychiatry. 2005;162:475–484. Sokunbi MO, Gradin VB, Waiter GD, Cameron GG, Ahearn TS, Murray AD, et al. Nonlinear complexity analysis of brain FMRI signals in schizophrenia. PloS One. 2014;9:e95146. Xue S-W, Yu Q, Guo Y, Song D, Wang Z. Resting-state brain entropy in schizophrenia. Compr Psychiatry. 2019;89:16–21. Yang AC, Hong C-J, Liou Y-J, Huang K-L, Huang C-C, Liu M-E, et al. Decreased resting-state brain activity complexity in schizophrenia characterized by both increased regularity and randomness. Hum Brain Mapp. 2015;36:2174–2186. van Erp TGM, Hibar DP, Rasmussen JM, Glahn DC, Pearlson GD, Andreassen OA, et al. Subcortical brain volume abnormalities in 2028 individuals with schizophrenia and 2540 healthy controls via the ENIGMA consortium. Mol Psychiatry. 2016;21:547–553. Okada N, Fukunaga M, Yamashita F, Koshiyama D, Yamamori H, Ohi K, et al. Abnormal asymmetries in subcortical brain volume in schizophrenia. Mol Psychiatry. 2016;21:1460–1466. Liang J, Chen L, Li Y, Chen Y, Yuan L, Qiu Y, et al. Unraveling the Prefrontal Cortex-Basolateral Amygdala Pathway’s Role on Schizophrenia’s Cognitive Impairments: A Multimodal Study in Patients and Mouse Models. Schizophr Bull. 2024;50:913–923. Haber SN. Corticostriatal circuitry. Dialogues Clin Neurosci. 2016;18:7–21. Loonen AJM, Ivanova SA. New insights into the mechanism of drug-induced dyskinesia. CNS Spectr. 2013;18:15–20. Perez-Rando M, García-Martí G, Escarti MJ, Salgado-Pineda P, McKenna PJ, Pomarol-Clotet E, et al. Alterations in the volume and shape of the basal ganglia and thalamus in schizophrenia with auditory hallucinations. Prog Neuropsychopharmacol Biol Psychiatry. 2024;131:110960. Avram M, Brandl F, Bäuml J, Sorg C. Cortico-thalamic hypo- and hyperconnectivity extend consistently to basal ganglia in schizophrenia. Neuropsychopharmacol Off Publ Am Coll Neuropsychopharmacol. 2018;43:2239–2248. Giraldo-Chica M, Rogers BP, Damon SM, Landman BA, Woodward ND. Prefrontal-Thalamic Anatomical Connectivity and Executive Cognitive Function in Schizophrenia. Biol Psychiatry. 2018;83:509–517. Wagner G, De la Cruz F, Schachtzabel C, Güllmar D, Schultz CC, Schlösser RG, et al. Structural and functional dysconnectivity of the fronto-thalamic system in schizophrenia: a DCM-DTI study. Cortex J Devoted Study Nerv Syst Behav. 2015;66:35–45. Xue S-W, Yu Q, Guo Y, Song D, Wang Z. Resting-state brain entropy in schizophrenia. Compr Psychiatry. 2019;89:16–21. Freedman R, Olsen-Dufour AM, Olincy A, Consortium on the Genetics of Schizophrenia. P50 inhibitory sensory gating in schizophrenia: analysis of recent studies. Schizophr Res. 2020;218:93–98. Xia L, Yuan L, Du X-D, Wang D, Wang J, Xu H, et al. P50 inhibition deficit in patients with chronic schizophrenia: Relationship with cognitive impairment of MATRICS consensus cognitive battery. Schizophr Res. 2020;215:105–112. Lang X, Wang D, Zhou H, Wang L, Kosten TR, Zhang X-Y. P50 inhibition defects, psychopathology and gray matter volume in patients with first-episode drug-naive schizophrenia. Asian J Psychiatry. 2023;80:103421. Xia L, Wang D, Wei G, Wang J, Zhou H, Xu H, et al. P50 inhibition defects with psychopathology and cognitive impairment in patients with first-episode drug naïve schizophrenia. Prog Neuropsychopharmacol Biol Psychiatry. 2021;107:110246. Sumich A, Harris A, Flynn G, Whitford T, Tunstall N, Kumari V, et al. Event-related potential correlates of depression, insight and negative symptoms in males with recent-onset psychosis. Clin Neurophysiol Off J Int Fed Clin Neurophysiol. 2006;117:1715–1727. Coull JT. Neural correlates of attention and arousal: insights from electrophysiology, functional neuroimaging and psychopharmacology. Prog Neurobiol. 1998;55:343–361. Crowley KE, Colrain IM. A review of the evidence for P2 being an independent component process: age, sleep and modality. Clin Neurophysiol Off J Int Fed Clin Neurophysiol. 2004;115:732–744. Boutros NN, Korzyukov O, Jansen B, Feingold A, Bell M. Sensory gating deficits during the mid-latency phase of information processing in medicated schizophrenia patients. Psychiatry Res. 2004;126:203–215. Folstein JR, Van Petten C. Influence of cognitive control and mismatch on the N2 component of the ERP: a review. Psychophysiology. 2008;45:152–170. Oribe N, Hirano Y, Kanba S, del Re E, Seidman L, Mesholam-Gately R, et al. Progressive Reduction of Visual P300 Amplitude in Patients With First-Episode Schizophrenia: An ERP Study. Schizophr Bull. 2015;41:460–470. Bahramali H, Gordon E, Li WM, Rennie C, Wright J, Meares R. Fast and slow reaction times and associated ERPs in patients with schizophrenia and controls. Int J Neurosci. 1998;95:155–165. Muller-Gass A, Macdonald M, Schröger E, Sculthorpe L, Campbell K. Evidence for the auditory P3a reflecting an automatic process: elicitation during highly-focused continuous visual attention. Brain Res. 2007;1170:71–78. Atkinson RJ, Michie PT, Schall U. Duration mismatch negativity and P3a in first-episode psychosis and individuals at ultra-high risk of psychosis. Biol Psychiatry. 2012;71:98–104. Hermens DF, Ward PB, Hodge MAR, Kaur M, Naismith SL, Hickie IB. Impaired MMN/P3a complex in first-episode psychosis: cognitive and psychosocial associations. Prog Neuropsychopharmacol Biol Psychiatry. 2010;34:822–829. Takahashi H, Rissling AJ, Pascual-Marqui R, Kirihara K, Pela M, Sprock J, et al. Neural substrates of normal and impaired preattentive sensory discrimination in large cohorts of nonpsychiatric subjects and schizophrenia patients as indexed by MMN and P3a change detection responses. NeuroImage. 2013;66:594–603. Mayer AR, Hanlon FM, Franco AR, Teshiba TM, Thoma RJ, Clark VP, et al. The neural networks underlying auditory sensory gating. NeuroImage. 2009;44:182–189. Knott V, Millar A, Fisher D. Sensory gating and source analysis of the auditory P50 in low and high suppressors. NeuroImage. 2009;44:992–1000. Tregellas JR, Davalos DB, Rojas DC, Waldo MC, Gibson L, Wylie K, et al. Increased hemodynamic response in the hippocampus, thalamus and prefrontal cortex during abnormal sensory gating in schizophrenia. Schizophr Res. 2007;92:262–272. Mulert C, Jäger L, Schmitt R, Bussfeld P, Pogarell O, Möller H-J, et al. Integration of fMRI and simultaneous EEG: towards a comprehensive understanding of localization and time-course of brain activity in target detection. NeuroImage. 2004;22:83–94. Tables Table 1. Sample Characteristics and Treatment Outcomes (Mean ± SD). Variable Baseline (n=90) 8 weeks follow up (n=90) P -value Age (years) 25..99±9.23 Education (years) 11.34±3.49 Gender (male/female) 36/44 Age of onset (years) 23.93±9.432 Duration(years) 2.24±1.87 PANSS-T 126.10±20.57 60.00±24.49 P <0.0001 PANSS-P 26.02±7.65 12.04±5.33 P <0.0001 PANSS-N 34.44±6.75 17.63±7.85 P <0.0001 PANSS-G 65.63±11.10 30.32±13.13 P <0.0001 Note: SD, standard deviation; PANSS, Positive and Negative Syndrome Scale; PANSS-T, PANSS total scores; PANSS-P, PANSS positive symptom scores; PANSS-N, PANSS negative symptom scores; PANSS-G, PANSS general psychopathological symptom scores; Group comparisons were performed using the Wilcoxon signed-rank test. Additional Declarations The authors have declared there is NO conflict of interest to disclose Supplementary Files SupplementaryMaterials.docx Supplementary Materials FigureS1.tif Figure S1 Cite Share Download PDF Status: Posted Version 1 posted 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-6047108","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":452206658,"identity":"66d29564-7c89-4def-87ce-bd49f5958d59","order_by":0,"name":"Xiang-Yang Zhang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA40lEQVRIiWNgGAWjYDACCQglByKYSdJiTLqWxAaitcjPbn728Gvb4fTt/GcMPxcw2MkzsJ89gFcL45xj5sYyZw7n7pyRYyw9gyHZsIEnLwGvFmaJBDNpiYrDuRtu8Jgx8zAwJzBI8Bjg1cImkf5NWsLgcLrB+TMgLfWEtfBI5JhJfqg4nGBwIAek5TBhLRISOWXSDGfSDXfOSCuW5jE4btjGk4Nfi/yM9G2SP9us5c35D2/8zFNRLc/Pfga/FhAAuqeZwYCBA6gSiNgIqgcCxh8MdUDF7A+IUTwKRsEoGAUjEAAAen46PELQVmgAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0003-3326-382X","institution":"Affiliated Mental Health Center of Anhui Medical University","correspondingAuthor":true,"prefix":"","firstName":"Xiang-Yang","middleName":"","lastName":"Zhang","suffix":""},{"id":452206659,"identity":"357bf664-2b56-44b9-ba9f-37491cd74ce3","order_by":1,"name":"Liju Liu","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Liju","middleName":"","lastName":"Liu","suffix":""},{"id":452206660,"identity":"358b5a9a-c2ca-447b-8128-a71c126d34e6","order_by":2,"name":"Dongmei Wang","email":"","orcid":"","institution":"Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Dongmei","middleName":"","lastName":"Wang","suffix":""},{"id":452206661,"identity":"ed51f951-7bb4-4d3e-b8ce-694edc393e03","order_by":3,"name":"Meng Chen","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Meng","middleName":"","lastName":"Chen","suffix":""},{"id":452206662,"identity":"d42db2a5-7294-4fba-aab1-26faeb12ada1","order_by":4,"name":"Xiaoe Lang","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Xiaoe","middleName":"","lastName":"Lang","suffix":""},{"id":452206663,"identity":"94d7749f-7980-4fc8-9405-64cc6ad7755a","order_by":5,"name":"Mi Yang","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Mi","middleName":"","lastName":"Yang","suffix":""}],"badges":[],"createdAt":"2025-02-17 10:47:38","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6047108/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6047108/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":82560163,"identity":"a6427a6b-3504-4896-a978-372dd638f9d3","added_by":"auto","created_at":"2025-05-13 01:32:54","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":5709246,"visible":true,"origin":"","legend":"\u003cp\u003ePrediction of PANSS-T score. \u0026nbsp;The left side of each panel presents the correlations between the predicted and observed reduction of PANSS-T score across individuals, and the right side of each panel shows the features selected at least five times from the 10 CV iterations in prediction of symptom changes. \u0026nbsp;A. Prediction of PANSS-T score with WPE features as predictors. B. Prediction of core PANSS-T score with EEG features as predictors. C. Prediction of PANSS-T score with WPE and EEG features as predictors. Features in orange were positively correlated with reduction of symptom, while features in blue were negatively correlated with reduction of symptom.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-6047108/v1/d5b4696348db6194ddc15b10.png"},{"id":82559106,"identity":"d5b97e07-e599-4cbc-b3f9-2d7aa661d936","added_by":"auto","created_at":"2025-05-13 01:24:54","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":3642915,"visible":true,"origin":"","legend":"\u003cp\u003ePrediction of PANSS-N score. \u0026nbsp;The left side of each panel presents the correlations between the predicted and observed reduction of PANSS-N score across individuals, and the right side of each panel shows the features selected at least five times from the 10 CV iterations in prediction of symptom changes. \u0026nbsp;A. Prediction of core PANSS-N score with WPE features as predictors. B. Prediction of core PANSS-N score with EEG features as predictors. C. Prediction of core PANSS-N score with WPE and EEG features as predictors. Features in orange were positively correlated with reduction of symptom, while features in blue were negatively correlated with reduction of symptom.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-6047108/v1/9ce8931638a7c392c27d3c63.png"},{"id":82559096,"identity":"ba3c7fc3-e1e0-4eb7-b1c4-152a93d2023c","added_by":"auto","created_at":"2025-05-13 01:24:53","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":5002882,"visible":true,"origin":"","legend":"\u003cp\u003ePrediction of PANSS-G score. \u0026nbsp;The left side of each panel presents the correlations between the predicted and observed reduction of PANSS-G score across individuals, and the right side of each panel shows the features selected at least five times from the 10 CV iterations in prediction of symptom changes. \u0026nbsp;A. Prediction of core PANSS-G score with WPE features as predictors. B. Prediction of core PANSS-G score with EEG features as predictors. C. Prediction of core PANSS-G score with WPE and EEG features as predictors. Features in orange were positively correlated with reduction of symptom, while features in blue were negatively correlated with reduction of symptom.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-6047108/v1/22f5dc9c410c8f01b8489ec5.png"},{"id":82559103,"identity":"506bf342-1b54-470d-aeba-245f036c9813","added_by":"auto","created_at":"2025-05-13 01:24:54","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1735850,"visible":true,"origin":"","legend":"\u003cp\u003eCorrelation between WPE and EEG features.\u003cstrong\u003e \u003c/strong\u003eHeatmap showing the Pearson correlations between the top WPE and EEG features (selected across all 10 outer folds). A significant correlation is observed between the WPE of the right amygdala and the P50 ratio (* \u003cem\u003eP\u003c/em\u003e\u0026lt;0.05), while no significant correlations are found for other features.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-6047108/v1/f2c0190c0db88caa10045daf.png"},{"id":84687512,"identity":"d9112c05-f1d5-4c54-9b88-db2fada0c39e","added_by":"auto","created_at":"2025-06-16 09:09:02","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":17107682,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6047108/v1/42bd3a7e-4cd6-4d1e-9f55-4d31a3079f57.pdf"},{"id":82560161,"identity":"238ba1ac-671e-4bd0-9281-8c58bc8c9985","added_by":"auto","created_at":"2025-05-13 01:32:53","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":589946,"visible":true,"origin":"","legend":"Supplementary Materials","description":"","filename":"SupplementaryMaterials.docx","url":"https://assets-eu.researchsquare.com/files/rs-6047108/v1/426d2ece53e203654d3595cd.docx"},{"id":82560165,"identity":"50024870-5f3c-4362-9234-6dfef0acbf6a","added_by":"auto","created_at":"2025-05-13 01:32:54","extension":"tif","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":35497376,"visible":true,"origin":"","legend":"Figure S1","description":"","filename":"FigureS1.tif","url":"https://assets-eu.researchsquare.com/files/rs-6047108/v1/75766f9f251d313298c1d2cd.tif"}],"financialInterests":"The authors have declared there is \u003cb\u003eNO\u003c/b\u003e conflict of interest to disclose","formattedTitle":"Enhancing Prediction of Individualized Antipsychotic Outcome with fMRI-EEG Feature Integration in First-Episode Schizophrenia: A Real-World Study","fulltext":[{"header":"Introduction","content":"\u003cp\u003eSchizophrenia, as a severe mental disorder, presents with a wide range of clinical manifestations, including cognitive dysfunction, emotional instability, and social impairment[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Although antipsychotic medication treatment remains the primary treatment approach for schizophrenia, the response to medication varies significantly among patients[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e], which complicates the development of personalized treatment plans. This challenge is particularly pronounced in real-world settings, where treatment outcomes are influenced by various factors, including individual physiological characteristics, the diversity of therapeutic approaches, and changes in the external environment. Therefore, accurately predicting the effects of personalized pharmacotherapy under conditions more reflective of clinical practice is crucial. In recent years, the rapid advancement of neuroimaging techniques has provided important tools for predicting treatment efficacy in schizophrenia. Among these, functional magnetic resonance imaging (fMRI) and electroencephalography (EEG), as complementary neuroimaging modalities, reveal different aspects of brain information processing from spatial and temporal dimensions, respectively, laying the foundation for exploring the neurobiological mechanisms underlying treatment prediction.\u003c/p\u003e \u003cp\u003efMRI, with its high spatial resolution, captures functional connectivity between brain regions, resting-state network properties, and spatial patterns of local brain activity. For example, Cao et al. and Sarpal et al. found that functional connectivity in the superior temporal cortex and the striatum could predict the treatment response to antipsychotic medications in first-episode drug-naive (FEDN) schizophrenia patients[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Blessing et al. reported that anterior hippocampal-cortical functional connectivity could distinguish patients from healthy controls and predict the response to second-generation antipsychotics[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Li et al. observed increased fractional amplitude of low frequency fluctuation in the left ventral striatum in FEDN schizophrenia patients across two independent samples, which predicted individual treatment response to olanzapine[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Additionally, a previous study of ours found reduced complexity in the functional activity of the striatum in FEDN schizophrenia patients, with baseline values significantly correlating with clinical symptom improvement after treatment with risperidone[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. These studies indicate that fMRI can capture spatial pattern features of brain function, providing important clues for treatment prediction.\u003c/p\u003e \u003cp\u003eEEG, with its high temporal resolution, captures dynamic changes in neurophysiological activity during brain information processing and has been successfully employed in several studies to predict patients' responses to clozapine treatment[\u003cspan additionalcitationids=\"CR9\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Event-related potentials (ERPs), an important component of EEG, serve as sensitive neurophysiological markers for real-time monitoring of brain responses to information processing. Among these, the reduction in P300 amplitude and the delay in latency are characteristic features of schizophrenia patients, which not only worsen with disease progression[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e] but also track changes in clinical symptoms[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. P50, a key indicator of sensory gating ability, reflects the brain's capacity to filter irrelevant information at the perceptual level. Nagamoto et al. found that P50 gating ability was closely related to the response to clozapine treatment[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e], with atypical antipsychotics outperforming traditional antipsychotics in enhancing P50 gating[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. These findings suggest that EEG holds promising potential for reflecting dynamic neural activity and treatment response in schizophrenia patients.\u003c/p\u003e \u003cp\u003eAlthough both fMRI and EEG have shown certain advantages in unimodal efficacy prediction studies, single-modality data often fail to comprehensively reflect the complex neurobiological mechanisms of schizophrenia. In real-world settings, the diversity of treatment regimens (including medication types and dosages) further complicates prediction. Atypical antipsychotic drugs, due to their lower side effects and broader applicability[\u003cspan additionalcitationids=\"CR18\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e], particularly their significant effects on improving cognitive and emotional symptoms[\u003cspan additionalcitationids=\"CR21 CR22\" citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e], have become first-line treatment options. However, no study has yet integrated fMRI and EEG features to predict the individualized treatment response to atypical antipsychotics in FEDN schizophrenia patients within real-world settings. Therefore, there is an urgent need to explore whether the combination of these two modalities can significantly enhance the accuracy of efficacy prediction.\u003c/p\u003e \u003cp\u003eIn this study, we investigate whether combined fMRI and EEG features can predict individualized clinical outcomes in FEDN schizophrenia patients in real-world settings and assess the potential improvement in predictive accuracy compared to unimodal features. The study incorporates personalized treatment regimens without restrictions on medication types or dosages, thus more comprehensively reflecting the complexity of real-world treatment and enhancing the clinical applicability of the model. By integrating multimodal features across both spatial and temporal dimensions, this research provides a theoretical basis for uncovering the neurobiological mechanisms underlying treatment response in schizophrenia patients and for advancing the clinical translation of multimodal imaging technologies.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eParticipants\u003c/h2\u003e \u003cp\u003eThis study included 90 FEDN schizophrenia outpatients recruited from the First Hospital of Shanxi Medical University. The inclusion criteria were as follows: 1) met the DSM-IV diagnostic criteria for schizophrenia, with the diagnosis established by two trained psychiatrists based on structured clinical interview for DSM-IV (SCID)[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]; 2) Positive and Negative Syndrome Scale (PANSS) score[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]\u0026thinsp;\u0026ge;\u0026thinsp;60; 3) no prior treatment with psychotropic medications; 4) age\u0026thinsp;\u0026lt;\u0026thinsp;45 years, Han nationality; and 5) illness duration no longer than 5 years. The exclusion criteria were: 1) presence of metabolic diseases such as diabetes and hyperlipidemia, organic brain diseases, or a history of traumatic brain injury; 2) hearing or visual impairments that could affect brain evoked potential examinations; and 3) other psychiatric disorders or substance dependence (except for nicotine) according to DSM-IV criteria. All participants provided written informed consent, and the study protocol was approved by the Institutional Review Board of the First Hospital of Shanxi Medical University.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eMedication and Clinical Assessments\u003c/h3\u003e\n\u003cp\u003eAll patients received second-generation antipsychotic medications (e.g., olanzapine, risperidone, or aripiprazole) as prescribed by their attending physicians based on individual clinical conditions, without the inclusion of psychotherapy or other non-pharmacological interventions. The selection of treatment regimens and dosage adjustments were made according to the patient's condition, drug tolerance, and the clinical judgment of the physician. All patients underwent a second clinical assessment between the 6th and 8th week after the initiation of treatment. Clinical symptoms were assessed using the 30-item PANSS.\u003c/p\u003e\n\u003ch3\u003efMRI Data Acquisition and Preprocessing\u003c/h3\u003e\n\u003cp\u003eResting-state fMRI data were acquired on a General Electric Signa HDxt 3.0 T scanner with a 32-channel head coil at baseline. Functional data were obtained using a gradient-echo echo-planar imaging sequence [TR/TE\u0026thinsp;=\u0026thinsp;2000ms/30ms, flip angle\u0026thinsp;=\u0026thinsp;90\u0026deg;, field of view\u0026thinsp;=\u0026thinsp;240\u0026times;240 mm\u003csup\u003e2\u003c/sup\u003e, matrix size\u0026thinsp;=\u0026thinsp;64 \u0026times; 64; slice thickness\u0026thinsp;=\u0026thinsp;4mm, voxel size\u0026thinsp;=\u0026thinsp;3.75 \u0026times; 3.75 \u0026times; 4 mm\u003csup\u003e3\u003c/sup\u003e, 210 time points collected].\u003c/p\u003e \u003cp\u003eThe fMRI data preprocessing was performed using the Data Processing Assistant for Resting-State fMRI (DPARSF, V5.4, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://rfmri.org/DPARSF\u003c/span\u003e\u003cspan address=\"http://rfmri.org/DPARSF\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). To minimize magnetization equilibrium effects, the first 10 volumes were discarded. Slice timing correction and realignment were applied to the remaining images. The functional images were then normalized to the EPI template with a resampled voxel size of 3mm \u0026times; 3mm \u0026times; 3mm. Several covariates, including Friston 24 motion parameters, cerebrospinal fluid, and white matter signals, were regressed out as nuisance variables to reduce spurious variance. Spatial smoothing (FWHM\u0026thinsp;=\u0026thinsp;6mm), detrending, and band-pass filtering (0.01\u0026ndash;0.08 Hz) were applied to the data. Participants were excluded from further analysis if their translational or rotational displacement exceeded 3.0 mm or 3.0\u0026deg;, or if their mean frame-wise displacement (a measure of micromovement) exceeded 0.3 mm. Twelve participants were excluded due to excessive head motion, leaving 78 participants for the final feature extraction and prediction tasks. The demographic and clinical characteristics of participants included for analysis are shown in supplementary Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e.\u003c/p\u003e\n\u003ch3\u003eCalculation of WPE Features\u003c/h3\u003e\n\u003cp\u003eWeighted Permutation Entropy (WPE) is a measure of time series complexity that quantifies the unpredictability and irregularity of the signal. It calculates the degree of randomness in the data by examining the order of values within the time series. Higher WPE values indicate more complex and unpredictable dynamics, while lower values suggest more regular patterns. In this study, regional BOLD time series were extracted using the Anatomical Automatic Labeling (AAL) atlas[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e], which defines 90 distinct cerebral functional regions, excluding the cerebellum. The WPE for each brain region was then calculated based on its entire time series. Detailed information on the WPE calculation method can be found in our previous studies[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e].\u003c/p\u003e\n\u003ch3\u003eParadigm\u003c/h3\u003e\n\u003cp\u003eThe P50 components were measured using a conditioning-testing paradigm. Participants were instructed to relax and focus on a fixed point. We generated a stimulus signal of 90-dB pulses of 0.1ms in duration and recorded the event-related potential waveforms. A total of 32 pairs of auditory stimuli were presented, with an inter-stimulus interval of 500ms and an inter-pair interval of 10s[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. Each epoch lasted for 1000ms, including a 100ms before stimulation 1, 500ms between stimuli and 400ms after stimulation 2.\u003c/p\u003e \u003cp\u003eThe N100, P200, N200, and P300 components were measured using an auditory oddball paradigm. Participants were presented with frequent standard tones (500 Hz, 80%), infrequent target tones (1000 Hz, 10%), and infrequent novel distractor sounds (variety of sounds, 10%). Standard and target tones were 50 milliseconds in duration (5-millisecond rise/fall time) at an 80-dB sound pressure level. Stimuli were presented with a 1.25-second stimulus onset asynchrony. Participants were instructed to press a response button with their preferred hand upon hearing the target stimuli, while ignoring the standard and novel sounds.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eEEG Data Acquisition and Preprocessing\u003c/h2\u003e \u003cp\u003eA signal generator and a digital 40-channel EBNeuro Sirius EEG system (EBNeuro, Florence, Italy) were used for electrophysiological recordings and EEG signal processing. EEG signals were filtered using a 0.1\u0026ndash;250 Hz analog filter and a 50 Hz notch filter and sampled at a frequency of 500 Hz. Eye movements were recorded using Ag-AgCl disc electrodes to capture electro-oculography (EOG) signals, and ocular artifacts were corrected offline using the independent component analysis method. Vertical EOG signals were recorded with electrodes placed above and below the left eye, while horizontal EOG signals were recorded with electrodes placed at the outer canthi of both eyes. The ground electrode was positioned at the midpoint between FPz and Fz, and the reference electrode was placed on the right mastoid. Offline correction of reference voltage was conducted using the algebraic average of the left and right mastoids. The impedance of all electrodes was maintained below 5kΩ. Trials with artifacts exceeding\u0026thinsp;\u0026plusmn;\u0026thinsp;50\u0026micro;V were excluded. Offline analysis was conducted using bandpass filtering (10\u0026ndash;50 Hz, 12 dB/octave roll-off). To control background noise during the experiment, continuous 70 dB(A) broadband white noise was provided throughout the recording.\u003c/p\u003e \u003cp\u003eFor the P50 component, S1 and S2 amplitudes and latencies were extracted from the Cz electrode. The S1 amplitude was defined as the maximum peak within the 40\u0026ndash;80ms time window following the S1 stimulus onset, while the S2 amplitude was determined at the latency window closest to the S1 latency after the S2 stimulus. The P50 ratio, an index of sensory gating, was calculated by dividing the S2 amplitude by the S1 amplitude.\u003c/p\u003e \u003cp\u003eFor the N100, P200, N200, and P300 components, individual peak amplitudes and latencies were assessed within the following time windows: N100 (60-180ms), P200 (150-250ms), N200 (200-350ms), and P300 (230-420ms). The amplitudes and latencies of N100, P200, and N200 components were extracted from the Cz electrode. The P3b peak amplitudes and latencies were chosen from the target minus standard difference wave as the most positive peak at electrode Pz (where P3b is maximum), while P3a peak amplitudes and latencies were identified from the novel minus standard difference wave at electrode Cz (where P3a is maximum). All ERP components were extracted by two independent neurophysiologists who were blinded to the participants' status. In cases of discrepancies in the analysis results, a third party conducted further evaluations to ensure consistency and accuracy.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003ePrediction of Treatment Outcome with WPE and EEG Measures\u003c/h3\u003e\n\u003cp\u003eThis study aims to investigate whether EEG features and WPE features can predict changes in PANSS scores for patients with schizophrenia under real-world settings. Using the LASSO regression model combined with cross-validation, EEG features (15), WPE features (90 regions), and their combination were used as predictors to assess their ability to predict changes in the PANSS total score, positive, negative, and general symptom subscales (PANSS scores at baseline minus PANSS scores at follow-up). We also compared the predictive performance of single-modal (EEG or WPE) and multi-modal features. All features were standardized before entering the model to eliminate scale differences and ensure comparisons on the same scale. Additionally, to control for potential confounding factors, feature values were adjusted for covariates, including patient sex, age, and illness duration, prior to standardization. LASSO regression is an L1-norm regularization method that introduces a shrinkage penalty term (λ) to avoid overfitting and automatically shrinks the coefficients of less important features to zero. It is particularly suitable for datasets with high multicollinearity. Model evaluation was performed using a repeated nested cross-validation method (10 outer folds and 10 inner folds). In each outer fold, the inner folds were used to optimize λ, and predictions were made on the test set. After cross-validation, the predicted reduction scores for each subject were compared to the actual reduction scores. Model performance was evaluated by the correlation between predicted and observed values, and significance was determined using 1000 permutations. In each permutation, subject labels were randomly shuffled, and the entire cross-validation procedure was repeated on the shuffled data. The P-value was calculated as the proportion of correlation coefficients in the null models greater than the observed ones.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003ePrediction of PANSS-T Score\u003c/h2\u003e \u003cp\u003eAs shown in Figure. 1A, using WPE features as predictors, the cross-validated LASSO regression revealed significant correlations between predicted and actual reductions in the PANSS-T score (\u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.288, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.003). The selected features for the model are listed on the right. Specifically, the top features contributing to symptom changes (features selected across all 10 outer folds) included the bilateral orbital part of middle frontal gyrus, right triangular part of inferior frontal gyrus, left supplementary motor area, left medial orbital part of superior frontal gyrus, bilateral amygdala, left calcarine, right lingual, right inferior occipital gyrus, right superior parietal gurus, bilateral supramarginal gyrus, left precuneus, bilateral caudate, right thalamus and right temporal pole.\u003c/p\u003e \u003cp\u003eAs shown in Figure. 1B, using EEG features as predictors, the cross-validated LASSO regression revealed significant correlations between predicted and actual reductions in the PANSS-T score (\u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.401, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.002). Note that the result remained highly significant (\u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.344, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.001) even after removing an outlier at the top right of the scatter plot, as shown in supplementary Figure. S1A. The top features contributing to symptom changes included the amplitude of P200, latency of N200, amplitude of P3b and P50 ratio.\u003c/p\u003e \u003cp\u003eAs shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC, the LASSO regression model using the combination of WPE and EEG features predicted the PANSS total score with a significant correlation (\u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.440, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001). The top features contributing to symptom changes included the left middle frontal gyrus, bilateral orbital part of middle frontal gyrus, right triangular part of inferior frontal gyrus, right amygdala, left calcarine, right inferior occipital gyrus, right supramarginal gyrus, bilateral caudate, right thalamus, left Heschl gyrus, latency of N100, amplitude of P200, latency of N200 and P50 ratio.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003ePrediction of PANSS-N Score\u003c/h2\u003e \u003cp\u003eAs shown in Figure. 2A, using WPE features as predictors, the cross-validated LASSO regression revealed significant correlations between predicted and actual reductions in the PANSS-N score (\u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.232, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.021). The top features contributing to symptom changes included the left anterior cingulate and paracingulate gyrus, left middle occipital gyrus, right inferior occipital gyrus, right supramarginal gyrus, bilateral caudate and left pallidum.\u003c/p\u003e \u003cp\u003eAs shown in Figure. 2B, using EEG features as predictors, the cross-validated LASSO regression revealed significant correlations between predicted and actual reductions in the PANSS-N score (\u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.288, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.008). The top features contributing to symptom changes included the amplitude of P200, latency of N200 and latency of P3a. However, the result was no longer significant after excluding the outlier in the top-right corner.\u003c/p\u003e \u003cp\u003eAs shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC, the LASSO regression model using the combination of WPE and EEG features predicted the PANSS-N score with a significant correlation (\u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.328, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.002). The top features contributing to symptom changes included the left middle occipital gyrus, right supramarginal gyrus, left caudate left pallidum, amplitude of P200 and latency of P3a.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003ePrediction of PANSS-P Score\u003c/h2\u003e \u003cp\u003eNeither single-modality features nor the combined features effectively predicted the improvement in positive symptoms after treatment. Using WPE features, the prediction result was \u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.129, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.148; with EEG features, the result was \u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.121, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.147; and with combined features, the prediction result was \u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.165, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.074, which is close to statistical significance but does not reach a significant level.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003ePrediction of PANSS-G Score\u003c/h2\u003e \u003cp\u003eAs shown in Figure. 3A, using WPE features as predictors, the cross-validated LASSO regression revealed significant correlations between predicted and actual reductions in the PANSS-G score (\u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.293, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.002). Note that the result remained highly significant (\u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.267, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) even after removing an outlier at the top right of the scatter plot, as shown in supplementary Figure. S1B. The top features contributing to symptom changes include the middle frontal gyrus, left orbital part of middle frontal gyrus, left supplementary motor area, right amygdala, right inferior occipital gyrus, right superior parietal gyrus, right inferior parietal, left caudate and right temporal pole.\u003c/p\u003e \u003cp\u003eAs shown in Figure. 3B, using EEG features as predictors, the cross-validated LASSO regression revealed significant correlations between predicted and actual reductions in the PANSS-N score (\u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.393, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.001). Note that the result remained highly significant (\u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.378, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) even after removing an outlier at the top right of the scatter plot, as shown in supplementary Figure. S1C. The top features contributing to symptom changes included the amplitude of P200, latency of N200, and P50 ratio.\u003c/p\u003e \u003cp\u003eAs shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC, the LASSO regression model using the combination of WPE and EEG features predicted the PANSS-N score with a significant correlation (\u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.449, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001). The top features contributing to symptom changes included the left\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eMiddle frontal gyrus, left orbital part of middle frontal gyrus, left supplementary motor area, right amygdala, left cuneus, right inferior occipital gyrus, right superior parietal gyrus, left caudate, latency of N200, and P50 ratio.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eCorrelation between WPE and EEG Features\u003c/h2\u003e \u003cp\u003eIn this study, we further investigated the correlation between WPE features and EEG features. Specifically, we extracted the top features selected in the LASSO model (features selected across all 10 outer folds) and computed the Pearson correlation between these WPE features and EEG features. The results showed that, except for the significant correlation between the WPE feature of the right amygdala and the P50 ratio (\u003cem\u003eR\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.290, \u003cem\u003eP\u003c/em\u003e\u003csub\u003euncorrected\u003c/sub\u003e=0.010), no significant correlations were found between the other features, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e. This result suggests that the improvement in the performance of the fusion model is not due to the correlation or redundancy between the two modal features, but rather because they provide complementary information that reflects different aspects of brain information processing.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study found that baseline fMRI and EEG features can effectively predict the improvement of psychiatric symptoms, including negative symptoms, general symptoms, and total symptom scores, in FEDN schizophrenia patients after 6\u0026ndash;8 weeks of treatment with atypical antipsychotics. The multimodal fusion model demonstrated higher predictive capability compared to unimodal features, as evidenced by an improvement in the model's R-value and the lack of significant correlations between key features selected in the fusion model. This suggests that these features provide complementary information in reflecting the brain's information processing mechanisms. The selected key features include WPE characteristics from brain regions such as the middle frontal gyrus, triangular part of the inferior frontal gyrus, supplementary motor area, amygdala, caudate, pallidum, and thalamus, as well as latency of N100, N200, P3a, amplitude of P200, and P50 ratio, further revealing the potential roles of these features in predicting treatment outcomes.\u003c/p\u003e \u003cp\u003eThis study has several notable advantages: First, the participants included were all DNFE schizophrenia patients, ensuring that the fMRI and EEG features could more purely reflect the pathophysiological characteristics of schizophrenia itself, without the confounding effects of disease progression or previous pharmacological treatments. Second, this study was conducted in a real-world setting, where treatment regimens were not restricted by medication types or dosages, encompassing a variety of atypical antipsychotic treatments. This design more closely mirrors clinical practice, enhancing the model's generalizability and clinical applicability. Third, unlike previous studies that have focused on group-level binary classification of antipsychotic treatment response, this study emphasizes individualized efficacy prediction. Group-level responder/non-responder classification methods have certain limitations, as they rely on arbitrary classification criteria and overlook intra-group variability in treatment effects, which may complicate individualized treatment decisions for clinicians. In contrast, this study overcomes these limitations by predicting treatment efficacy as a continuous variable, providing a model with greater clinical reference value. Fourth, we are the first to validate the superiority and complementarity of combined EEG and fMRI features in efficacy prediction. By combining the high temporal resolution of EEG with the high spatial resolution of fMRI, this study revealed the synergistic effect of the two modalities across different dimensions of information processing, further enhancing the predictive power of treatment responses in schizophrenia patients and providing valuable support for the application of multimodal imaging technologies in schizophrenia research.\u003c/p\u003e \u003cp\u003eOur study found that the complexity of brain regions, including the middle frontal gyrus, triangular part of the inferior frontal gyrus, supplementary motor area, amygdala, caudate, pallidum, and thalamus, significantly contributes to predicting clinical symptom changes after antipsychotic treatment. In schizophrenia-related research, brain activity complexity, as an indicator of information processing capacity[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e], has provided unique insights into the neurobiological mechanisms of the disorder. The middle frontal gyrus and triangular part of the inferior frontal gyrus are critical components of the prefrontal cortex, and dysfunction in these areas is closely associated with symptoms such as impaired executive function, working memory deficits[\u003cspan additionalcitationids=\"CR32\" citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e], and disorganization[\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. Numerous studies have reported abnormal complexity in the frontal lobe of schizophrenia patients, including increased BOLD signal complexity in the left inferior frontal gyrus and Brodmann area 47[\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e], as well as decreased complexity in the middle and superior frontal regions[\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]. These findings collectively suggest a disruption in the information processing abilities of the prefrontal cortex. Research has shown that schizophrenia patients exhibit reduced amygdala volume[\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e], which correlates with lower cognitive function scores; mouse models indicate that activation of the prefrontal cortex-basolateral amygdala pathway can improve cognitive performance[\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e], suggesting that amygdala dysfunction plays a significant role in cognitive deficits in schizophrenia. The caudate, pallidum, and thalamus form a crucial part of the cortico-striato-pallido-thalamo-cortical circuit[\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e], which plays a central role in regulating cognitive functions and behavioral responses. Abnormalities in this circuit are thought to underlie the symptoms and cognitive deficits of schizophrenia[\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. In a previous study, we found that the complexity of brain activity in the caudate nucleus, putamen, and pallidum was significantly reduced in first-episode schizophrenia patients, and the complexity of the caudate nucleus was closely related to cognitive deficits such as attention maintenance and verbal fluency[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. On the other hand, the thalamus, as a hub for information transmission and sensory integration, has low connectivity with the prefrontal cortex, which is associated with impairments in higher cognitive functions such as attention and working memory[\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e]. Xue et al. reported a significant reduction in the complexity of the bilateral thalamus in schizophrenia patients[\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e], suggesting that the dynamic information processing capacity of the thalamus may be impaired, which could affect the filtering and integration of multisensory information, further exacerbating cognitive deficits. In summary, the predictive model in this study successfully captured the key brain regions associated with functional abnormalities in schizophrenia, providing important biological evidence for individualized efficacy prediction.\u003c/p\u003e \u003cp\u003eOur study also found that the latency of N100, N200, and P3a, the amplitude of P200, and the P50 ratio all significantly contributed to predicting clinical symptom changes following antipsychotic treatment. These ERP components reflect the neurophysiological activity at different stages of information processing in the brain, revealing multi-level impairments in perception, attention, and cognitive functions in schizophrenia patients. P50 gating deficits are a hallmark feature of schizophrenia[\u003cspan additionalcitationids=\"CR49\" citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e], reflecting an impairment in sensory information filtering. In a previous study, we found a significant correlation between P50 gating and general psychopathology as well as PANSS total scores in first-episode schizophrenia patients[\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e], further supporting its value in predicting treatment response. Next, N100, an early response to sensory input, with reduced amplitude and delayed latency, indicates dysfunction in the auditory cortex and early deficits in selective attention in schizophrenia patients[\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e]. A meta-analysis showed that most studies reported reduced P200 amplitude in schizophrenia patients, while a few showed increased or no difference[\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e]. Changes in P200 amplitude further highlight potential abnormalities in stimulus encoding and selective attention in patients. These attention-related abnormalities are not limited to the early stages of information processing but may extend to subsequent sensory gating processes related to attention[\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e], thereby affecting higher-order information processing efficiency. At later stages of cognitive processing, abnormalities in N200 and P300 are more prominent. N200 is primarily evoked by the frontal and central brain regions and is closely related to cognitive control, novelty processing, and matching processes[\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e]. Reduced amplitude and delayed latency of N200 suggest impairments in conflict monitoring and cognitive flexibility in patients[\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e, \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e]. P3a mainly reflects the attentional shift toward task-irrelevant stimuli or information[\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e]. Several studies have found that the auditory P3a response elicited in passive auditory oddball tasks is attenuated in psychotic-spectrum disorders and those at risk for psychosis[\u003cspan additionalcitationids=\"CR61\" citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e]. These results suggest that the predictive model in our study successfully captured the multi-level abnormalities in perception, attention, and cognitive processing reflected by ERP components, validating their predictive value as neurophysiological biomarkers and providing support for the development of individualized treatment plans.\u003c/p\u003e \u003cp\u003eIt is noteworthy that our study found no significant correlation between the fMRI and EEG features extracted by the fusion model, suggesting that the independence and complementarity of these two modalities in reflecting different dimensions of information are key to enhancing the model's predictive performance. Several EEG-fMRI joint studies have shown that brain regions such as the thalamus, middle frontal gyrus, and dorsolateral prefrontal cortex are involved in P50 gating[\u003cspan additionalcitationids=\"CR64\" citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e]. Tregellas et al. found that P50 gating ability was significantly correlated with BOLD signals in the thalamus[\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e], highlighting its important role in sensory information filtering. P300, an important ERP component associated with target detection and attentional shifting, is considered a marker of processing speed and efficiency, with its amplitude and latency reflecting these aspects. Horovitz et al. showed that in oddball tasks, the probability of target events not only influenced P300 amplitude and latency but also synchronously affected BOLD signal changes in related brain regions, such as the supramarginal gyri, thalamus, insula and right medial frontal gyrus. Additionally, Mulert et al. found consistent activations of fMRI and EEG signals in most regions during target detection tasks, including the temporo-parietal junction, supplementary motor area/anterior cingulate cortex, insula, and middle frontal gyrus[\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e]. These results emphasize the synergistic role of EEG and fMRI in cognitive processes. The advantages of the fusion model lie not only in the complementarity of temporal and spatial resolution but also in capturing the synergistic effects of static and dynamic information processing. EEG features reflect information filtering and target detection on a short time scale, while WPE indices reveal the complexity of functional brain activity, particularly in resource regulation and dynamic adaptation. This complementarity enables the fusion model to more comprehensively characterize information processing abnormalities in schizophrenia patients and combine the neurobiological mechanisms of temporal and spatial dimensions, providing important support for the efficiency and accuracy of the predictive model.\u003c/p\u003e \u003cp\u003eThe present study has some limitations. First, First, although the inclusion of DNFE schizophrenia patients minimizes the confounding effects of prior treatment, the relatively small sample size, particularly for a machine learning study focusing on individual prediction, may limit the generalizability of our findings to broader clinical populations. Second, although this study focused on treatment prediction in real-world settings, the lack of detailed information on the specific types and doses of antipsychotic medications prevented adjustment for these potential confounding factors before incorporating features into the model, which may impact the accuracy and interpretability of the predictions. Third, this study lacked an independent dataset to validate the performance improvements of the fusion model, which may limit its generalizability across different samples and clinical settings.\u003c/p\u003e \u003cp\u003eIn conclusion, this study provides initial evidence that the integration of fMRI-based complexity measures and EEG-derived ERP components enhances the prediction of antipsychotic treatment outcomes in DNFE schizophrenia patients within real-world settings. These findings underscore the advantage of multi-modal neuroimaging in capturing complementary information across spatial and temporal dimensions, offering valuable insights for advancing precision psychiatry and individualized treatment strategies.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eSupplementary Information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSupplementary information is available at MP's website.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData are available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research was funded by National Natural Science Foundation of China (62373079),\u0026nbsp;Science and Technology Department of Sichuan Province\u0026nbsp;(2024ZYD0039),\u0026nbsp;Health Commission of Sichuan Province(24CXTD11),\u0026nbsp;Sichuan Medical Association (S23012),\u0026nbsp;Chengdu Science and Technology Bureau (2022-YF05-01867-SN),\u0026nbsp;Health Commission of Chengdu\u0026nbsp;(2024141),\u0026nbsp;CAS International Cooperation Research Program (153111KYSB20190004) and STI2030-Major Projects 2021ZD0202102.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eC\u003c/strong\u003e\u003cstrong\u003eompeting\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;Interest\u003c/strong\u003e\u003cstrong\u003es\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCRediT\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eA\u003c/strong\u003e\u003cstrong\u003euthorship\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eC\u003c/strong\u003e\u003cstrong\u003eontribution\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eS\u003c/strong\u003e\u003cstrong\u003etatement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLiju Liu:\u003c/strong\u003e Methodology, Conceptualization, Visualization, Writing-original draft\u003cstrong\u003e. Dongmei Wang\u003c/strong\u003e\u003cstrong\u003e:\u003c/strong\u003e Supervision, Writing—review \u0026amp; editing. \u003cstrong\u003eMeng Chen:\u0026nbsp;\u003c/strong\u003eFormal analysis, Writing-original draft.\u003cstrong\u003eXiaoE Lang\u003c/strong\u003e\u003cstrong\u003e:\u003c/strong\u003e Investigation, Writing—review \u0026amp; editing.\u0026nbsp;\u003cstrong\u003eMi Yang:\u003c/strong\u003e Resources, Writing-original draft, Writing-review \u0026amp; editing.\u0026nbsp;\u003cstrong\u003eXiangyang Zhang:\u003c/strong\u003e Investigation, Resources, Writing-review \u0026amp; editing.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eMueser KT, McGurk SR. Schizophrenia. The Lancet. 2004;363:2063\u0026ndash;2072.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrenner HD, Dencker SJ, Goldstein MJ, Hubbard JW, Keegan DL, Kruger G, et al. Defining treatment refractoriness in schizophrenia. Schizophr Bull. 1990;16:551\u0026ndash;561.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCao B, Cho RY, Chen D, Xiu M, Wang L, Soares JC, et al. Treatment response prediction and individualized identification of first-episode drug-na\u0026iuml;ve schizophrenia using brain functional connectivity. Mol Psychiatry. 2020;25:906\u0026ndash;913.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSarpal DK, Argyelan M, Robinson DG, Szeszko PR, Karlsgodt KH, John M, et al. Baseline Striatal Functional Connectivity as a Predictor of Response to Antipsychotic Drug Treatment. Am J Psychiatry. 2016;173:69\u0026ndash;77.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlessing EM, Murty VP, Zeng B, Wang J, Davachi L, Goff DC. Anterior Hippocampal-Cortical Functional Connectivity Distinguishes Antipsychotic Na\u0026iuml;ve First-Episode Psychosis Patients From Controls and May Predict Response to Second-Generation Antipsychotic Treatment. Schizophr Bull. 2020;46:680\u0026ndash;689.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi H, Guo W, Liu F, Chen J, Su Q, Zhang Z, et al. Enhanced baseline activity in the left ventromedial putamen predicts individual treatment response in drug-naive, first-episode schizophrenia: Results from two independent study samples. EBioMedicine. 2019;46:248\u0026ndash;255.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu L, Li Z, Kong D, Huang Y, Wu D, Zhao H, et al. Neuroimaging markers of aberrant brain activity and treatment response in schizophrenia patients based on brain complexity. Transl Psychiatry. 2024;14:1\u0026ndash;12.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKhodayari-Rostamabad A, Hasey GM, Maccrimmon DJ, Reilly JP, de Bruin H. A pilot study to determine whether machine learning methodologies using pre-treatment electroencephalography can predict the symptomatic response to clozapine therapy. Clin Neurophysiol Off J Int Fed Clin Neurophysiol. 2010;121:1998\u0026ndash;2006.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRavan M, Hasey G, Reilly JP, MacCrimmon D, Khodayari-Rostamabad A. A machine learning approach using auditory odd-ball responses to investigate the effect of Clozapine therapy. Clin Neurophysiol Off J Int Fed Clin Neurophysiol. 2015;126:721\u0026ndash;730.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMasychev K, Ciprian C, Ravan M, Manimaran A, Deshmukh A. Quantitative biomarkers to predict response to clozapine treatment using resting EEG data. Schizophr Res. 2020;223:289\u0026ndash;296.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMathalon DH, Ford JM, Rosenbloom M, Pfefferbaum A. P300 reduction and prolongation with illness duration in schizophrenia. Biol Psychiatry. 2000;47:413\u0026ndash;427.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFord JM, Mathalon DH, Marsh L, Faustman WO, Harris D, Hoff AL, et al. P300 amplitude is related to clinical state in severely and moderately ill patients with schizophrenia. Biol Psychiatry. 1999;46:94\u0026ndash;101.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMathalon DH, Ford JM, Pfefferbaum A. Trait and state aspects of p300 amplitude reduction in schizophrenia: a retrospective longitudinal study. Biol Psychiatry. 2000;47:434\u0026ndash;449.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNagamoto HT, Adler LE, McRae KA, Huettl P, Cawthra E, Gerhardt G, et al. Auditory P50 in schizophrenics on clozapine: improved gating parallels clinical improvement and changes in plasma 3-methoxy-4-hydroxyphenylglycol. Neuropsychobiology. 1999;39:10\u0026ndash;17.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLight GA, Geyer MA, Clementz BA, Cadenhead KS, Braff DL. Normal P50 Suppression in Schizophrenia Patients Treated With Atypical Antipsychotic Medications. Am J Psychiatry. 2000;157:767\u0026ndash;771.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBecker J, Gomes I, Ghisolfi ES, Schuch A, Ramos FLP, Ehlers JA, et al. Clozapine, but not typical antipsychotics, correct P50 suppression deficit in patients with schizophrenia. Clin Neurophysiol Off J Int Fed Clin Neurophysiol. 2004;115:396\u0026ndash;401.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeucht S, Corves C, Arbter D, Engel RR, Li C, Davis JM. Second-generation versus first-generation antipsychotic drugs for schizophrenia: a meta-analysis. Lancet Lond Engl. 2009;373:31\u0026ndash;41.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFabrazzo M, Cipolla S, Camerlengo A, Perris F, Catapano F. Second-Generation Antipsychotics\u0026rsquo; Effectiveness and Tolerability: A Review of Real-World Studies in Patients with Schizophrenia and Related Disorders. J Clin Med. 2022;11:4530.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeucht S, Priller J, Davis JM. Antipsychotic Drugs: A Concise Review of History, Classification, Indications, Mechanism, Efficacy, Side Effects, Dosing, and Clinical Application. Am J Psychiatry. 2024;181:865\u0026ndash;878.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBilder RM, Goldman RS, Volavka J, Czobor P, Hoptman M, Sheitman B, et al. Neurocognitive effects of clozapine, olanzapine, risperidone, and haloperidol in patients with chronic schizophrenia or schizoaffective disorder. Am J Psychiatry. 2002;159:1018\u0026ndash;1028.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGood KP, Kiss I, Buiteman C, Woodley H, Rui Q, Whitehorn D, et al. Improvement in cognitive functioning in patients with first-episode psychosis during treatment with quetiapine: an interim analysis. Br J Psychiatry Suppl. 2002;43:s45-49.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVeselinović T, Scharpenberg M, Heinze M, Cordes J, M\u0026uuml;hlbauer B, Juckel G, et al. Disparate effects of first and second generation antipsychotics on cognition in schizophrenia - Findings from the randomized NeSSy trial. Eur Neuropsychopharmacol J Eur Coll Neuropsychopharmacol. 2019;29:720\u0026ndash;739.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTorrisi SA, Laudani S, Contarini G, De Luca A, Geraci F, Manag\u0026ograve; F, et al. Dopamine, Cognitive Impairments and Second-Generation Antipsychotics: From Mechanistic Advances to More Personalized Treatments. Pharm Basel Switz. 2020;13:365.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFirst M, Spitzer R, Gibbon M, Williams J. Structured clinical interview for DSM-IV-TR Axis I Disorders, Research Version, Non-patient Edition. SCID-IP, 2002.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKay SR, Fiszbein A, Opler LA. The positive and negative syndrome scale (PANSS) for schizophrenia. Schizophr Bull. 1987;13:261\u0026ndash;276.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTzourio-Mazoyer N, Landeau B, Papathanassiou D, Crivello F, Etard O, Delcroix N, et al. Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. NeuroImage. 2002;15:273\u0026ndash;289.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang XY, Liu L, Liu S, Hong X, Chen DC, Xiu MH, et al. Short-Term Tropisetron Treatment and Cognitive and P50 Auditory Gating Deficits in Schizophrenia. Am J Psychiatry. 2012;169:974\u0026ndash;981.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXia L, Wang D, Wang J, Xu H, Huo L, Tian Y, et al. Association of cognitive and P50 suppression deficits in chronic patients with schizophrenia. Clin Neurophysiol Off J Int Fed Clin Neurophysiol. 2020;131:725\u0026ndash;733.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcDonough IM, Nashiro K. Network complexity as a measure of information processing across resting-state networks: evidence from the Human Connectome Project. Front Hum Neurosci. 2014;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang DJJ, Jann K, Fan C, Qiao Y, Zang Y-F, Lu H, et al. Neurophysiological Basis of Multi-Scale Entropy of Brain Complexity and Its Relationship With Functional Connectivity. Front Neurosci. 2018;12.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePerlstein WM, Carter CS, Noll DC, Cohen JD. Relation of prefrontal cortex dysfunction to working memory and symptoms in schizophrenia. Am J Psychiatry. 2001;158:1105\u0026ndash;1113.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePerlstein WM, Dixit NK, Carter CS, Noll DC, Cohen JD. Prefrontal cortex dysfunction mediates deficits in working memory and prepotent responding in schizophrenia. Biol Psychiatry. 2003;53:25\u0026ndash;38.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVan Snellenberg JX, Girgis RR, Horga G, van de Giessen E, Slifstein M, Ojeil N, et al. Mechanisms of Working Memory Impairment in Schizophrenia. Biol Psychiatry. 2016;80:617\u0026ndash;626.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMacDonald AW, Carter CS, Kerns JG, Ursu S, Barch DM, Holmes AJ, et al. Specificity of prefrontal dysfunction and context processing deficits to schizophrenia in never-medicated patients with first-episode psychosis. Am J Psychiatry. 2005;162:475\u0026ndash;484.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSokunbi MO, Gradin VB, Waiter GD, Cameron GG, Ahearn TS, Murray AD, et al. Nonlinear complexity analysis of brain FMRI signals in schizophrenia. PloS One. 2014;9:e95146.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXue S-W, Yu Q, Guo Y, Song D, Wang Z. Resting-state brain entropy in schizophrenia. Compr Psychiatry. 2019;89:16\u0026ndash;21.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang AC, Hong C-J, Liou Y-J, Huang K-L, Huang C-C, Liu M-E, et al. Decreased resting-state brain activity complexity in schizophrenia characterized by both increased regularity and randomness. Hum Brain Mapp. 2015;36:2174\u0026ndash;2186.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evan Erp TGM, Hibar DP, Rasmussen JM, Glahn DC, Pearlson GD, Andreassen OA, et al. Subcortical brain volume abnormalities in 2028 individuals with schizophrenia and 2540 healthy controls via the ENIGMA consortium. Mol Psychiatry. 2016;21:547\u0026ndash;553.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOkada N, Fukunaga M, Yamashita F, Koshiyama D, Yamamori H, Ohi K, et al. Abnormal asymmetries in subcortical brain volume in schizophrenia. Mol Psychiatry. 2016;21:1460\u0026ndash;1466.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiang J, Chen L, Li Y, Chen Y, Yuan L, Qiu Y, et al. Unraveling the Prefrontal Cortex-Basolateral Amygdala Pathway\u0026rsquo;s Role on Schizophrenia\u0026rsquo;s Cognitive Impairments: A Multimodal Study in Patients and Mouse Models. Schizophr Bull. 2024;50:913\u0026ndash;923.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHaber SN. Corticostriatal circuitry. Dialogues Clin Neurosci. 2016;18:7\u0026ndash;21.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLoonen AJM, Ivanova SA. New insights into the mechanism of drug-induced dyskinesia. CNS Spectr. 2013;18:15\u0026ndash;20.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePerez-Rando M, Garc\u0026iacute;a-Mart\u0026iacute; G, Escarti MJ, Salgado-Pineda P, McKenna PJ, Pomarol-Clotet E, et al. Alterations in the volume and shape of the basal ganglia and thalamus in schizophrenia with auditory hallucinations. Prog Neuropsychopharmacol Biol Psychiatry. 2024;131:110960.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAvram M, Brandl F, B\u0026auml;uml J, Sorg C. Cortico-thalamic hypo- and hyperconnectivity extend consistently to basal ganglia in schizophrenia. Neuropsychopharmacol Off Publ Am Coll Neuropsychopharmacol. 2018;43:2239\u0026ndash;2248.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGiraldo-Chica M, Rogers BP, Damon SM, Landman BA, Woodward ND. Prefrontal-Thalamic Anatomical Connectivity and Executive Cognitive Function in Schizophrenia. Biol Psychiatry. 2018;83:509\u0026ndash;517.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWagner G, De la Cruz F, Schachtzabel C, G\u0026uuml;llmar D, Schultz CC, Schl\u0026ouml;sser RG, et al. Structural and functional dysconnectivity of the fronto-thalamic system in schizophrenia: a DCM-DTI study. Cortex J Devoted Study Nerv Syst Behav. 2015;66:35\u0026ndash;45.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXue S-W, Yu Q, Guo Y, Song D, Wang Z. Resting-state brain entropy in schizophrenia. Compr Psychiatry. 2019;89:16\u0026ndash;21.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFreedman R, Olsen-Dufour AM, Olincy A, Consortium on the Genetics of Schizophrenia. P50 inhibitory sensory gating in schizophrenia: analysis of recent studies. Schizophr Res. 2020;218:93\u0026ndash;98.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXia L, Yuan L, Du X-D, Wang D, Wang J, Xu H, et al. P50 inhibition deficit in patients with chronic schizophrenia: Relationship with cognitive impairment of MATRICS consensus cognitive battery. Schizophr Res. 2020;215:105\u0026ndash;112.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLang X, Wang D, Zhou H, Wang L, Kosten TR, Zhang X-Y. P50 inhibition defects, psychopathology and gray matter volume in patients with first-episode drug-naive schizophrenia. Asian J Psychiatry. 2023;80:103421.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXia L, Wang D, Wei G, Wang J, Zhou H, Xu H, et al. P50 inhibition defects with psychopathology and cognitive impairment in patients with first-episode drug na\u0026iuml;ve schizophrenia. Prog Neuropsychopharmacol Biol Psychiatry. 2021;107:110246.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSumich A, Harris A, Flynn G, Whitford T, Tunstall N, Kumari V, et al. Event-related potential correlates of depression, insight and negative symptoms in males with recent-onset psychosis. Clin Neurophysiol Off J Int Fed Clin Neurophysiol. 2006;117:1715\u0026ndash;1727.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCoull JT. Neural correlates of attention and arousal: insights from electrophysiology, functional neuroimaging and psychopharmacology. Prog Neurobiol. 1998;55:343\u0026ndash;361.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCrowley KE, Colrain IM. A review of the evidence for P2 being an independent component process: age, sleep and modality. Clin Neurophysiol Off J Int Fed Clin Neurophysiol. 2004;115:732\u0026ndash;744.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoutros NN, Korzyukov O, Jansen B, Feingold A, Bell M. Sensory gating deficits during the mid-latency phase of information processing in medicated schizophrenia patients. Psychiatry Res. 2004;126:203\u0026ndash;215.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFolstein JR, Van Petten C. Influence of cognitive control and mismatch on the N2 component of the ERP: a review. Psychophysiology. 2008;45:152\u0026ndash;170.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOribe N, Hirano Y, Kanba S, del Re E, Seidman L, Mesholam-Gately R, et al. Progressive Reduction of Visual P300 Amplitude in Patients With First-Episode Schizophrenia: An ERP Study. Schizophr Bull. 2015;41:460\u0026ndash;470.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBahramali H, Gordon E, Li WM, Rennie C, Wright J, Meares R. Fast and slow reaction times and associated ERPs in patients with schizophrenia and controls. Int J Neurosci. 1998;95:155\u0026ndash;165.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMuller-Gass A, Macdonald M, Schr\u0026ouml;ger E, Sculthorpe L, Campbell K. Evidence for the auditory P3a reflecting an automatic process: elicitation during highly-focused continuous visual attention. Brain Res. 2007;1170:71\u0026ndash;78.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAtkinson RJ, Michie PT, Schall U. Duration mismatch negativity and P3a in first-episode psychosis and individuals at ultra-high risk of psychosis. Biol Psychiatry. 2012;71:98\u0026ndash;104.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHermens DF, Ward PB, Hodge MAR, Kaur M, Naismith SL, Hickie IB. Impaired MMN/P3a complex in first-episode psychosis: cognitive and psychosocial associations. Prog Neuropsychopharmacol Biol Psychiatry. 2010;34:822\u0026ndash;829.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTakahashi H, Rissling AJ, Pascual-Marqui R, Kirihara K, Pela M, Sprock J, et al. Neural substrates of normal and impaired preattentive sensory discrimination in large cohorts of nonpsychiatric subjects and schizophrenia patients as indexed by MMN and P3a change detection responses. NeuroImage. 2013;66:594\u0026ndash;603.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMayer AR, Hanlon FM, Franco AR, Teshiba TM, Thoma RJ, Clark VP, et al. The neural networks underlying auditory sensory gating. NeuroImage. 2009;44:182\u0026ndash;189.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKnott V, Millar A, Fisher D. Sensory gating and source analysis of the auditory P50 in low and high suppressors. NeuroImage. 2009;44:992\u0026ndash;1000.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTregellas JR, Davalos DB, Rojas DC, Waldo MC, Gibson L, Wylie K, et al. Increased hemodynamic response in the hippocampus, thalamus and prefrontal cortex during abnormal sensory gating in schizophrenia. Schizophr Res. 2007;92:262\u0026ndash;272.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMulert C, J\u0026auml;ger L, Schmitt R, Bussfeld P, Pogarell O, M\u0026ouml;ller H-J, et al. Integration of fMRI and simultaneous EEG: towards a comprehensive understanding of localization and time-course of brain activity in target detection. NeuroImage. 2004;22:83\u0026ndash;94.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003e\u003cstrong\u003eTable 1.\u0026nbsp;\u003c/strong\u003eSample Characteristics and Treatment Outcomes\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e(Mean\u0026nbsp;\u0026plusmn; SD).\u003c/p\u003e\n\u003cdiv\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"491\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003eVariable\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003eBaseline (n=90)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003e8 weeks follow up\u003c/p\u003e\n \u003cp\u003e(n=90)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cem\u003eP\u003c/em\u003e-value\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003eAge (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e25..99\u0026plusmn;9.23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003eEducation (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e11.34\u0026plusmn;3.49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003eGender (male/female)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e36/44\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003eAge of onset (years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e23.93\u0026plusmn;9.432\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003eDuration(years)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e2.24\u0026plusmn;1.87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003ePANSS-T\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e126.10\u0026plusmn;20.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003e60.00\u0026plusmn;24.49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003ePANSS-P\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e26.02\u0026plusmn;7.65\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003e12.04\u0026plusmn;5.33\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003ePANSS-N\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e34.44\u0026plusmn;6.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003e17.63\u0026plusmn;7.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003ePANSS-G\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 104px;\"\u003e\n \u003cp\u003e65.63\u0026plusmn;11.10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 132px;\"\u003e\n \u003cp\u003e30.32\u0026plusmn;13.13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 85px;\"\u003e\n \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u0026lt;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eNote: SD, standard deviation; PANSS, Positive and Negative Syndrome Scale; PANSS-T, PANSS total scores; PANSS-P, PANSS positive symptom scores; PANSS-N, PANSS negative symptom scores; PANSS-G, PANSS general psychopathological symptom scores; Group comparisons were performed using the Wilcoxon signed-rank test.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6047108/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6047108/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003ePredicting treatment efficacy in schizophrenia within real-world settings is of great clinical significance but challenging. fMRI and EEG reveal distinct spatial and temporal characteristics of brain information processing. Therefore, developing predictive models that integrate spatiotemporal multimodal features and accommodate the complexities of real-world environments is crucial for achieving accurate treatment outcome predictions and personalized therapy. Ninety first-episode, drug-naive schizophrenia patients underwent fMRI and EEG at baseline and received 6\u0026ndash;8 weeks of single antipsychotic treatment in a naturalistic setting. Clinical symptoms were evaluated using the Positive and Negative Syndrome Scale (PANSS) at baseline and post-treatment. Entropy metrics reflecting information processing capacity were calculated from BOLD signals as fMRI features, while key ERP components (P50, N100, P200, N200, and P300) representing different cognitive stages were extracted with their amplitudes and latencies as EEG features. LASSO regression model was used to assess the predictive power of unimodal and multimodal features for PANSS score reduction. The multimodal model outperformed unimodal models in predicting improvements in PANSS total score (R\u0026thinsp;=\u0026thinsp;0.440, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), negative symptom (R\u0026thinsp;=\u0026thinsp;0.328, P\u0026thinsp;=\u0026thinsp;0.002), and general psychopathology (R\u0026thinsp;=\u0026thinsp;0.449, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Key features in the multimodal model included WPE from the medial frontal gyrus, supplementary motor area, amygdala, caudate, pallidum, and thalamus, along with ERP characteristics like P50 ratio, N100, N200, P3a latencies, and P200 amplitude. These features were not significantly correlated, highlighting their complementary roles in information processing as key to the multimodal model's improved performance. This study demonstrates that multimodal fusion prediction model effectively integrates brain information processing features across different dimensions, significantly improving individualized prediction of treatment outcomes. The results hold important value for its translational application in precision psychiatry within real-world settings.\u003c/p\u003e","manuscriptTitle":"Enhancing Prediction of Individualized Antipsychotic Outcome with fMRI-EEG Feature Integration in First-Episode Schizophrenia: A Real-World Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-13 01:24:48","doi":"10.21203/rs.3.rs-6047108/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"0197b27b-efbc-4160-878c-961c650296fc","owner":[],"postedDate":"May 13th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":48092214,"name":"Health sciences/Biomarkers/Prognostic markers"},{"id":48092215,"name":"Health sciences/Diseases/Psychiatric disorders/Schizophrenia"}],"tags":[],"updatedAt":"2025-06-16T09:00:42+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-13 01:24:48","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6047108","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6047108","identity":"rs-6047108","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}