Levodopa-related physiomarkers during deep brain stimulation in Parkinson’s disease

Levodopa-related physiomarkers during deep brain stimulation in Parkinson’s disease | 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 Levodopa-related physiomarkers during deep brain stimulation in Parkinson’s disease Martijn G.J. de Neeling, Carina R. Oehrn, Mariëlle J. Stam, Bart J. Keulen, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9190790/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 13 You are reading this latest preprint version Abstract Adaptive deep brain stimulation (aDBS) can adjust stimulation to the (medication) state of patients with Parkinson’s disease using physiological biomarkers (physiomarkers), but whether a universal physiomarker frequency exists remains unknown. We quantified levodopa-induced spectral changes (2–100 Hz) during active DBS before and after levodopa intake in local field potential recordings from 52 PD patients (100 hemispheres). Using cluster-based permutation statistics, we assigned the frequency with the largest significant change as the personalized physiomarker. Alpha/low-beta (9–20 Hz) power reduction was the most common physiomarker (35% of hemispheres) followed by stimulation-entrained gamma (SEG, 62.5 Hz) power increases (19% of hemispheres). Strong levodopa responses, high levodopa doses, and lower stimulation amplitudes were mostly associated with alpha/low-beta physiomarkers, while higher stimulation amplitudes and average levodopa doses were associated with SEG physiomarkers. These findings argue against a single generalizable physiomarker frequency and support individualized selection based on patient-specific levodopa and stimulation profiles. Health sciences/Biomarkers Health sciences/Medical research Health sciences/Neurology Biological sciences/Neuroscience Figures Figure 1 Figure 2 Figure 3 Introduction Deep brain stimulation (DBS) of the subthalamic nucleus (STN) is an effective treatment for advanced Parkinson’s disease (PD)( 1 – 3 ). It is particularly beneficial for managing levodopa-related motor fluctuations, such as medication-induced dyskinesia and bradykinesia related to medication wearing off ( 4 ). Yet, patients receiving DBS may still experience remaining motor fluctuations, often due to stimulation-induced side effects that prevent further increases in stimulation amplitude ( 1 , 5 – 7 ). Therefore, adapting stimulation amplitude according to the disease state (e.g. higher stimulation when medication is wearing off) could improve the efficacy of DBS. Adaptive DBS (aDBS) automatically adjusts the stimulation amplitude based upon specific neurophysiological markers of symptom severity, referred to as physiomarkers ( 8 ). To date, the most established physiomarker is the spectral power within the beta (± 13–30 Hz) frequency band of STN local field potentials (LFPs) that captures the occurrence and strength of beta oscillations. This is based on several studies showing that, in the absence of stimulation and medication (STIM OFF, MED OFF), beta power is positively correlated with the severity of (contralateral) bradykinesia and rigidity ( 9 ) and reduced by therapeutic interventions such as medication and DBS ( 8 , 10 – 12 ). Nevertheless, the magnitude of beta power explains only ~ 17% of variability in symptom severity across patients ( 9 ). Furthermore, beta oscillations are co-modulated by voluntary movement, which complicates use of this physiomarker for aDBS ( 13 ). In line, a recent study (n = 4) found that STN beta power measured in ambulatory settings was only predictive of motor symptoms in one of six hemispheres ( 14 ). Recent studies have pointed out stimulation-entrained gamma (SEG) oscillations (typically at 62.5 Hz with 125 Hz stimulation frequency, a subharmonic) as an alternative physiomarker of symptom severity and medication state ( 14 – 16 ). A first aDBS feasibility study in the at-home environment demonstrated that SEG recorded in the STN and sensorimotor cortex using electrocorticography (ECoG), showed sharp increases after levodopa intake and decreases when levodopa is wearing off and was more informative of the patient’s medication state than beta power ( 14 ). Although this pilot study (n = 4) suggested SEG may be preferable over the use of beta power as physiomarker for aDBS, it is not yet known how many patients exhibit SEG activity in the STN or which factors contribute to its occurrence. In order to identify the optimal physiomarker related to the dopaminergic state during stimulation in a larger cohort, we investigated the effect of medication intake on the spectral power of LFP signals in a large cohort of 52 patients (100 hemispheres) using a data-driven analysis without a-priori assumption of the physiomarker frequency ( 14 ). In addition, we used a k-means clustering method to explain the occurrence and magnitude of the two most commonly found physiomarkers by the clinical characteristics of the patients. Methods Participants The STROBE criteria were used in the reporting of this study ( 17 ). Patients with Parkinson’s disease and bilateral DBS surgery were included between November 2022 and November 2024 as part of the AI-DBS study ( 18 ). SenSight directional leads (model 33005, Medtronic, Minneapolis, MN, US) were implanted in the subthalamic nucleus (STN) together with the Medtronic Percept ™ PC neurostimulator using previously described procedures followed by our center ( 19 ) in the context of standard clinical care. This study was approved by the local ethics committee (NL80384.018.22) and was carried out in accordance with the Declaration of Helsinki. Informed written consent was received from all patients. Exclusion criteria included the lack of clinical data or the (full) LFP recording at the follow-up time point after continuous DBS (cDBS) optimization, use of a continuous intrajejunal or subcutaneous (fos)levodopa, and serious adverse events after DBS surgery. Data acquisition The data described in this study were collected during the 6-month follow-up visit after DBS implementation. Variability in motor symptom severity was assessed by scoring the Movement Disorders Society Unified Parkinson’s Disease Rating Scale part III (MDS-UPDRS-III) while sensing LFP time series with BrainSense ™ Streaming in two conditions: ON stimulation and without medication (“STIM ON MED OFF”) and ON stimulation and with medication (STIM ON MED ON) (Fig. 1 ). The MED ON state was assessed 1 hour after intake of a suprathreshold (120%) dose of the morning levodopa equivalent daily dose (LEDD) with a minimum of 100 mg. Each patient started with the STIM ON MED OFF scoring. For each of the two conditions, we conducted a recording at rest for 1 minute and for 3 to 7 minutes while assessing the MDS-UPDRS-III score. LFPs were recorded using the surrounding contacts of the contact used for stimulation. The LFP signal was high-pass filtered at 1 Hz and low-pass filtered at 100 Hz, sampled at 250 Hz, and stored on the Percept ™ device. After the session, the data were retrieved, together with the stimulation amplitude, from the clinician programmer ( 20 ). The suprathreshold levodopa dose per patient was extracted from clinical records. We analyzed video recordings to determine the start- and end time of the minute resting state and the MDS-UPDRS-III scoring. Synchronization of video- and LFP data was performed by filming the onset of the recording prior to clinical examination. Data processing We used MATLAB (R2023b, The MathWorks, Inc., Natick,MA, USA) and the FieldTrip toolbox (version 1.0.2.0 ( 21 )) for our analyses. Patients with insufficient resting state data, defined by less than 30 seconds starting from the beginning of each resting state recording, were excluded from the analysis. We first utilised a single value decomposition method for suppressing ECG artifacts within the recordings ( 22 ). Then, we computed the power spectral density (PSD) of non-interrupted time segments based on all available data per hemisphere using Welch’s method, a window size of 1 second and 95% overlap, resulting in a PSD over 1-100 Hz with 1-Hz frequency bins. These single PSD estimates per hemisphere served to compare the MED OFF and MED ON state on a group level using all data available (without separation of rest and movement data). For this, each PSD was normalized by dividing each power value by the maximum value per STN. Subsequently, the median power and IQR was determined across hemispheres per frequency bin. We then repeated the PSD computation after dividing the available time series into non-overlapping epochs of 2 seconds duration, hence resulting in multiple PSD estimates per hemisphere. For each hemisphere, outlying power values exceeding five standard deviations were replaced by the median value across PSDs per frequency bin. The set of PSD estimates per hemisphere served to determine a personalized, data-driven physiomarker of the medication state. For this, we used unnormalized spectra and focused on the 2-100 Hz frequency range. These PSDs were computed separately for the resting state (30 seconds − 1 minute), movement state with a matched duration to the resting state, and the full duration of the MDS-UPDRS-III scoring (3–7 minutes of data). The duration-matched movement state was randomly selected from the total MDS-UPDRS-III scoring segment to make sure that detected physiomarkers were not based upon one specific movement task. Statistical analysis Physiomarker detection In order to determine personalised physiomarkers of the medication state, we used cluster-based non-parametric permutation tests ( 23 ) as implemented in the permutest.m ( 24 ) function (10.000 permutations, t-statistic, two-sided paired T-test, p < 0.05) to statistically compare OFF- and ON medication state PSDs for each hemisphere separately. We used the same permutation test for determining spectral changes from OFF- to ON MED across hemispheres (one PSD per hemisphere) to investigate the difference between within- and across-hemispheres results. We determined the magnitude of the statistical difference per hemisphere by calculating Cohen’s d effect size per 1-Hz bin and selected the frequency bin with the largest effect size within a statistically significant cluster as the “physiomarker”, similar to Oehrn et al. (Fig. 1 ) ( 14 ). This could result in multiple significant frequency bins per hemisphere in case multiple significant clusters were detected, however, only the largest effect size within the power spectrum from all significant clusters was selected as physiomarker. We regarded any effect size between − 0.2 and 0.2 too small and therefore non-significant, hence leaving the possibility of hemispheres with no detectable physiomarker ( 25 ). We then calculated the percentage of the total number of physiomarkers found (across all hemispheres) in the theta (4–8 Hz), alpha (9–12 Hz), low beta (13–20 Hz), high beta (21–30 Hz), low gamma (31–59 Hz), SEG (60–65 Hz with 125 Hz stimulation frequency ( 14 )) and high gamma (66–100 Hz) band to investigate in which frequency band most physiomarkers were detected. The percentage of physiomarkers between states was compared per frequency band with a Kruskal-Wallis test and post-hoc Mann-Whitney U test. Consistency of physiomarkers across hemispheres Next, we investigated the predominant direction of effect per frequency band across all personalized physiomarkers to test for similarity in presumed underlying neuronal mechanisms. Combined neighboring frequency bands were used for further analyses if this resulted in a common significant effect direction opposed to classic frequency band definitions in literature that were determined without active stimulation ( 8 ). Within-band effect directions were assessed by computing the mean sign of all effect directions across all physiomarkers within each band, i.e. discarding the magnitude of the effect size but only considering its directionality with + 1 indicating an increase in power after medication intake and − 1 a decrease. For each clinical state (rest, duration-matched movement, full MDS-UPDRS-III scoring), a two-tailed binomial test evaluated whether the observed proportion of MED ON > MED OFF versus MED OFF > MED ON directions deviated significantly from an equal 50/50 distribution. Mean sign values closer to 0 were interpreted as greater directional conflict, whereas mean sign closer to -1/1 were taken to indicate more consistent effects across hemispheres. Only the mean sign values for which the binomial test indicated a significant deviation from a 50/50 distribution were considered a significant, common effect direction per frequency band. To assess which of these three datasets (resting state, matched-duration movement state or the full MDS-UPDRS-III scoring) most consistently identified medication state, we averaged the mean sign values across frequency bands, and selected the dataset with the highest value (full MDS-UPDRS-III scoring) for all subsequent analyses. Clinical characteristics associated with physiomarker effect size and frequency We performed a clustering analysis to discover distinguishing sets of clinical characteristics associated with physiomarkers across hemispheres ( 26 ). Specifically, we tested for shared variance between several clinical variables and the magnitude of the physiomarker in the two frequency bands with most detected physiomarkers (i.e., alpha/low-beta and SEG). In hemispheres containing either the most common physiomarker (alpha/low-beta) or second most common physiomarker (SEG), for each hemisphere the mean effect size for all 1-Hz bins for each of these two frequency bands (alpha/low-beta and SEG) was combined with selected clinical features in a k-means clustering analysis to partition the data into groups based on their similarity. The clinical features included levodopa dose in mg, stimulation amplitude in milli-amperage, disease burden (MED OFF STIM OFF score), MDS-UPDRS-III response to levodopa during stimulation, and disease duration in years. The MDS-UPDRS-III response to levodopa during stimulation was defined as the STIM ON MED ON MDS-UPDRS-III hemibody score subtracted by the STIM ON MED OFF MDS-UPDRS-III hemibody score. The hemibody score was calculated by summing the rigidity, bradykinesia and tremor scores for either side of the body. The variables were pair-wise tested for collinearity (r > 0.5), which was the case for disease burden, the MDS-UPDRS-III response to levodopa during stimulation, and disease duration. Only the MDS-UPDRS-III response to levodopa during stimulation was retained as variable to avoid collinearity of variables that may distort cluster formation and bias results. Prior to clustering, all features were standardized (z-scored) to ensure equal weighting across dimensions. We used the Silhouette score to measure how similar each observation is to its own cluster compared to other clusters across a range of cluster numbers (two to eight clusters, Supplementary Figure S6) ( 27 ), which resulted in four clusters yielding the highest median score. The MATLAB function kmeans.m was used with the k-means + + initialization strategy to improve centroid placement and convergence. To reduce sensitivity to initialization and avoid convergence to local minima, the algorithm was repeated 100 times with different random initializations. The solution with the lowest within-cluster sum of squared distances was retained. To visualize the resulting clusters in a reduced-dimensional space, a principal component analysis (PCA) was applied to the feature matrix ( 28 ). The first three principal components, which together explained more than 70% of the variance, were plotted in 3-D scatter plots. Cluster separation was further highlighted by overlaying boundary outlines around each cluster and color-coding individual points according to their cluster assignment. To statistically evaluate differences between clusters, a one-way analysis of variance (ANOVA, p < 0.05) was performed on each feature. Post-hoc Tukey’s honestly significant difference (HSD, p < 0.05) test was then applied to identify which cluster pairs significantly differed from each other. Box plots per clinical variable including all data points separated by their respective cluster were plotted for visualization. For each cluster, we counted the number of hemispheres with a physiomarker detected in the alpha/low-beta and SEG frequency bands, to obtain insights into the clinical variables associated with physiomarker frequency. Discrimination of levodopa-related full spectral responses according to physiomarker frequency Finally, to investigate possible co-occurrence of multiple physiomarkers per hemisphere, we grouped the individual full power spectra (2-100 Hz) based on the frequency band of the detected physiomarker, and performed another cluster-based permutation test on the aggregated 2-second segmented PSD data. This data, without normalization, from all hemispheres per physiomarker frequency band (e.g. pooling all data from hemispheres that had an alpha/low-beta physiomarker (9–20 Hz)) was compared between MED ON and MED OFF and the median and IQR across segments was plotted separately per medication state. Results Clinical characteristics All data reported represent the median and interquartile range [IQR] unless otherwise specified. Data from fifty-two PD patients (31 male) with a total of 104 hemispheres (52 left, 52 right) were included in the study. Four hemispheres were excluded due to missing data, leaving 100 hemispheres (51 left, 49 right) for further analysis. Patients were 64 years old [59, 70] and had a disease duration of 10 [8, 14] years. During preoperative screening, the MDS-UPDRS-III MED OFF score was 43 [36, 53] and the MDS-UPDRS-III ON MED score was 18 [12, 25], showing an improvement after levodopa intake of 59% [51, 66%]. The MDS-UPDRS-III ON STIM MED OFF score after DBS titration (clinical follow up) was 23 [14, 30], showing a reduction of 49% [35, 50%] compared to the pre-surgical MDS-UPDRS-III score. The MDS-UPDRS-III ON STIM ON MED score at follow up was 19 [13, 23]. The levodopa equivalent daily dose (LEDD) during pre-operative screening was 1332 mg [935, 1674 mg] and 489 mg [300, 686 mg] at clinical follow up, showing a reduction of 60% [50, 75%]. Stimulation amplitude at follow up was 2.0 mA [1.6, 2.4 mA] and stimulation frequency was 125 Hz in all patients. A group analysis of PSDs across hemispheres (n = 100, 2-100 Hz) did not show a statistically significant difference between the STIM ON MED OFF phase compared to the STIM ON MED ON phase (p < 0.05, Supplementary Fig. 1). PSDs determined per hemisphere using the segmentation method (Fig. 1) resulted in a median of 183 (IQR [162, 232]) PSDs per hemisphere. Findings for the distribution of physiomarkers per dataset are summarized in Fig. 2. Distribution of physiomarkers during the resting state Two peaks were observed in the distribution of physiomarkers detected from the resting state LFP segments, one in the alpha/low-beta band and one in the SEG band (Fig. 2A; Table 1). For 13% of hemispheres the largest effect size of medication was within the alpha band, and in 18% of hemispheres in the low beta band. MED OFF > MED ON physiomarkers most commonly occurred at 14 Hz within the low beta band. For 11% of hemispheres the largest effect of medication was in the SEG band with a MED ON > MED OFF effect peak at 62 Hz. For 45% of hemispheres, the largest effect size was not found in the alpha, low beta or SEG band, but in the theta (15%), high beta (13%), low gamma (17%) or high gamma (13%) band. In general, physiomarkers were spread across a broad frequency range with only clear effect size directions in the theta, alpha-beta and SEG band (Fig. 2A). The median stimulation amplitude was 1.9 mA [1.5, 2.5 mA] for hemispheres with physiomarkers in the alpha-beta band and 2.3 mA [1.7, 2.6 mA] for hemispheres with physiomarkers in the SEG band without a significant difference (MWU test, p > 0.10). Table 1 – Percentage of hemispheres with detected physiomarkers per frequency band. Percentages are expressed compared to the total number of hemispheres in the analysis (n = 100) and listed separately for the three different states (resting state, movement, and full MDS-UPDRS-III scoring). The direction of the medication effect on LFP power of the physiomarker is specified between brackets, with ↓ illustrating a decrease in power when transitioning to the MED ON state and ↑ illustrating an increase in power when transitioning to the MED ON state. Low beta showed higher occurrence during full MDS-UPDRS-III scoring compared to the resting state (p = 0.03, indicated with an asterisk), while SEG physiomarkers showed a trend towards more occurrence between these two states (p = 0.08). A significant physiomarker was detected in all hemispheres (not significant = 0%). Frequency band Resting state (30 sec – 1 min) Movement (30 sec – 1 min) Full MDS-UPDRS-III scoring (3 min – 7 min) Theta (4–8 Hz) 15% (1% ↓, 14%↑) 9% (1% ↓, 8% ↑) 8% (0% ↓, 8% ↑) Alpha (9–12 Hz) 13% (10% ↓, 3% ↑) 8% (5% ↓, 3% ↑) 10% (7% ↓, 3% ↑) Low beta (13–20 Hz) 18% (17% ↓, 1% ↑) 20% (19% ↓, 1% ↑) 31% (28% ↓, 3% ↑)* High beta (21–30 Hz) 13% (3% ↓, 10% ↑) 15% (4% ↓, 11% ↑) 9% (2% ↓, 7% ↑) Low gamma (31–59 Hz) 17% (9% ↓, 8% ↑) 13% (3% ↓, 10% ↑) 10% (1% ↓, 9% ↑) SEG (60–65 Hz) 11% (1% ↓, 10% ↑) 17% (1% ↓, 16% ↑) 20% (1% ↓, 19% ↑) High gamma (66–100 Hz) 13% (7% ↓, 6% ↑) 18% (9% ↓, 9% ↑) 12% (5% ↓, 7% ↑) Not significant 0% 0% 0% Total ↓ vs. ↑ physiomarkers 48% / 52% 42% / 58% 44% / 56% Distribution of physiomarkers during the movement state For the duration-matched movement dataset, again two peaks were observed in the distribution of physiomarkers (Fig. 2B; Table 1) with 8% of hemispheres showing the largest effect size in the alpha band and 20% of hemispheres in the low beta band with a MED OFF > MED ON effect peak at 14 Hz. For 17% of hemispheres, the largest effect size was in the SEG band with a MED ON > MED OFF effect peak at 62 Hz. For 37% of hemispheres, the largest effect size was found in the theta (9%), high beta (15%), low gamma (13%) or high gamma (18%) band. There was an increased consistency of physiomarker effect directions per frequency band compared to the resting state with clear effect directions in the theta, low-beta and SEG band. A steep increase in the number of SEG physiomarkers was observed (Fig. 2B). Median stimulation amplitude was 1.9 mA [1.6, 2.4 mA] for hemispheres with physiomarkers in the alpha-beta band and 2.1 mA [1.7, 2.5 mA] for hemispheres with physiomarkers in the SEG band without a significant difference (MWU test, p > 0.10). Distribution of physiomarkers during the full MDS-UPDRS-III scoring To investigate the influence of data recording duration, we recomputed the distribution of physiomarkers based on LFP data acquired during the full MDS-UPDRS-III scoring (3–7 minutes), which showed two peaks in the alpha/low-beta and SEG band (Fig. 2C; Table 1). The largest effect of medication was seen in the alpha band for 10% of hemispheres and in the low beta band for 31% of hemispheres, with a MED OFF > MED ON effect peak at 13 Hz. In 35% of hemispheres a MED OFF > MED ON effect direction was seen in the alpha or low beta band. For 20% of hemispheres, the largest effect size was found in the SEG band with a MED ON > MED OFF effect peak at 62 Hz. For 27% of hemispheres, the largest effect size was not found in the alpha, low beta or SEG band, but in the theta (8%), high beta (9%), low gamma (10%) or high gamma (12%) band. Compared to the resting state and duration-match movement data, results for the full MDS-UPDRS-III scoring showed the most consistent effect directions across frequency bands with clear patterns emerging in the theta- (MED ON > MED OFF), alpha/low-beta- (MED OFF > MED ON), high-beta/low-gamma- (MED ON > MED OFF) and SEG- (MED ON > MED OFF) band (Fig. 2C, Supplementary Figure S2-S4). Using Kruskal-Wallis tests and post-hoc pairwise comparisons with MWU testing on all physiomarker occurrences per frequency band, low beta showed higher occurrence during full MDS-UPDRS-III scoring compared to the resting state (p = 0.03), while SEG physiomarkers showed a trend towards more occurrence between these two states (p = 0.08). Median stimulation amplitude was 1.9 mA [1.6, 2.4 mA] for hemispheres with physiomarkers in the alpha-beta band and 2.2 mA [1.7, 2.5 mA] for hemispheres with physiomarkers in the SEG band without a significant difference (MWU test). Consistency of physiomarkers across hemispheres The consistency of the effect size direction (e.g., predominantly MED OFF > MED ON or MED ON > MED OFF) was statistically tested for the physiomarkers in each frequency band for the resting state, movement state and during the full MDS-UPDRS-III scoring. A high value for the mean sign of the effect sizes (e.g., 1 when all hemispheres show a MED ON > MED OFF direction, -1 when all hemispheres show a MED OFF > MED ON direction; versus 0 when effects for the hemispheres are evenly split) was interpreted as a strong directional effect. Per state, binomial tests showed whether a frequency band had a statistically significant predominant effect direction (Table 2). Theta (MED ON > MED OFF), low beta (MED OFF > MED ON) and SEG (MED ON > MED OFF) showed a significant predominant effect direction in all states (p values are reported in Table 2). The alpha, high beta, and high gamma band showed no consistent effect direction in either state. For the low gamma band, a significant predominant effect direction (p = 0.02, MED ON > MED OFF) was found for the full MDS-UPDRS-III testing but not for the other states. When combining the high beta and low gamma band, a common effect direction was found (p = 0.00, MED ON > MED OFF) Table 2 – Predominance of physiomarker effect directions across hemispheres per frequency band. Mean sign values are reported based on the indicated number of hemispheres (n) for which a physiomarker was detected in the indicated frequency band or combination of frequency bands. Corresponding p-values of a binomial test indicate whether effect directions were significantly more often in the direction of an increase (↑) or a decrease (↓) in LFP power with levodopa or were not distinguishable from a 50/50 distribution (-). Findings are listed separately for the resting state, movement state, and full MDS-UPDRS-III scoring. Frequency band Mean sign (rest) Binomial test (p value) Mean sign (movement) Binomial test (p value) Mean sign (full MDS-UPDRS-III) Binomial test (p value) Theta (4–8 Hz) 0.87 (n = 15) 0.00* 0.78 (n = 9) 0.04* 1 (n = 8) 0.00* Alpha (9–12 Hz) -0.54 (n = 13) 0.09 -0.25 (n = 8) 0.73 0.4 (n = 10) 0.34 Low beta (13–20 Hz) -0.89 (n = 18) 0.00* -0.9 (n = 20) 0.00* -0.81 (n = 31) 0.00* High beta (21–30 Hz) 0.53 (n = 13) 0.09 0.47 (n = 15) 0.11 0.56 (n = 9) 0.18 Low gamma (31–59 Hz) -0.06 (n = 15) 1 0.54 (n = 13) 0.09 0.8 (n = 10) 0.02* SEG (60–65 Hz) 0.82 (n = 11) 0.01* 0.88 (n = 17) 0.00* 0.9 (n = 20) 0.00* High gamma (66–100 Hz) -0.08 (n = 13) 1 0.00 (n = 18) 1 0.17 (n = 12) 0.77 Alpha/low-beta (9–20 Hz) -0.74 (n = 31) 0.00* -0.71 (n = 28) 0.00* -0.71 (n = 41) 0.00* High beta/low gamma (21–59 Hz) 0.20 (n = 30) 0.36 0.5 (n = 28) 0.01* 0.68 (n = 19) 0.00* When combining the adjacent alpha and low beta band, a significant predominant effect direction was found. This can be explained by a general lower frequency of the beta peaks during active DBS as found in previously reported literature comparing the DBS OFF with the DBS ON state (8). Therefore, four distinct effects of medication on LFP power were observed: MED ON > MED OFF in theta and high-beta/low-gamma, MED OFF > MED ON in alpha/low-beta and MED ON > MED OFF in SEG. Alpha/low-beta and SEG were the frequency bands with most detected physiomarkers. Clinical characteristics associated with physiomarker effect size and frequency To investigate the association between clinical features and the presence of either an alpha/low-beta or SEG physiomarker across hemispheres and their magnitude of fluctuation, a k-means clustering analysis was performed using the mean of all 1-Hz bin effect sizes in the alpha/low beta and SEG band separately, stimulation amplitude (in mA), levodopa dose (in mg) and MDS-UPDRS-III response (in points reduction) to levodopa (five variables in total). Two hemispheres could not be included in the analysis due to incomplete MDS-UPDRS-III scores. The median silhouette score indicated a moderate structure for the optimum of four clusters (0.38, Supplementary Fig. 6) (27). The separation between the four clusters is visualized in Fig. 3 using three principal components that accounted for over 70% of the variance, a common method for visualizing clusters in a reduced-dimension space when multiple variables are used (28). For each cluster, we counted the number of hemispheres with physiomarkers in the alpha/low-beta and SEG band. Alpha/low-beta physiomarkers predominated cluster 1 and 3 (eleven out of twelve and eight out of nine hemispheres respectively), while SEG physiomarkers predominated cluster 2 (seven out of seven hemispheres). Cluster 4 contained a mixture of 14 hemispheres with alpha/low-beta physiomarkers and 10 hemispheres with SEG physiomarkers out of 24 hemispheres. Statistical evaluation using one-way ANOVA indicated that several features differed significantly across clusters (all p < 0.05, see Table 3). Pairwise testing with post-hoc Tukey’s HSD tests (Table 4) revealed a lower alpha/low-beta average effect size per hemisphere in cluster 1 compared to all other clusters and a lower alpha/low-beta average effect size per hemisphere in cluster 3 compared to cluster 2. Higher SEG average effect size per hemisphere was observed in cluster 2 compared to all other clusters. Cluster 3 had higher levodopa doses than cluster 2 and 4, cluster 1 contained lower stimulation amplitudes than cluster 2 and 3 and cluster 2 contained lower reductions in MDS-UPDRS-III scores following levodopa intake than cluster 1 and 4 but not compared to cluster 3. Broadly speaking, we can hence distinguish the following neurophysiological and clinical profiles: hemispheres in cluster 1 contained primarily alpha/low-beta physiomarkers, low stimulation amplitudes, average levodopa doses and a strong MDS-UPDRS-III response to levodopa; all hemispheres in cluster 2 contained SEG physiomarkers and showed relatively high stimulation amplitude, average levodopa dose and a weak MDS-UPDRS-III response to levodopa; hemispheres in cluster 3 contained primarily alpha/low-beta physiomarkers, high stimulation amplitudes, high doses and weak MDS-UPDRS-III responses to levodopa; cluster 4 (the largest group) contained a mixture of hemispheres with either alpha/low-beta and SEG physiomarkers and also showed the least distinctive clinical features. Stimulation amplitudes were relatively low compared cluster 3 (p = 0.02) and cluster 2 (p = 0.05). Levodopa doses were lower than cluster 2 and 3 (p = 0.00) but not compared to cluster 1, and MDS-UPDRS-III response to levodopa were weaker than cluster 1 (p = 0.00) but stronger than cluster 2 and 3 (p = 0.01, p = 0.01, respectively). Table 3 – Clinical characteristics of identified clusters and ANOVA results Average alpha/low-beta and SEG effect size per hemisphere, Stimulation amplitude (in mA), Dose (in mg), MDS-UPDRS-III response (in points reduction) were determined per cluster after k-means clustering. ANOVA analyses revealed significant differences in these metrics across clusters. Alpha/low-beta effect size per hemisphere (median, IQR) Cluster 1 Cluster 2 Cluster 3 Cluster 4 ANOVA (p-value) All clusters -0.57 [-0.71, -0.44] -0.01 [-0.22, 0.02] -0.36 [-0.51, -0.17] -0.17 [-0.28, -0.08] < 0.00 -0.26 [-0.44, -0.10] SEG effect size per hemisphere (median, IQR) 0.25 [0.13, 0.46] 0.97 [0.72, 1.29] 0.16 [-0.02, 0.29] 0.3 [0.14, 0.45] < 0.00 0.31 [0.14, 0.57] Stimulation (mA, median, IQR) 1.7 [1.4, 2.0] 2.5 [2.0, 2.6] 2.4 [2.1, 2.8] 1.9 [1.5, 2.3] < 0.00 2.0 [1.6, 2.4] Dose (mg, median, IQR) 150 [125, 200] 150 [150, 150] 200 [175, 350] 150 [100, 150] < 0.00 150 [150, 188] MDS-UPDRS-III response (points, median, IQR) -14 [-18, -11] -5 [-6, -4] -4 [-9, -3] -10 [-12, -8] < 0.00 -10 [-12, -6] Table 4 – Summary of post-hoc pair-wise statistical testing, using the Tukey HSD test. Alpha/low-beta effect size, SEG effect size, Stimulation, Dose, Levodopa MDS-UPDRS-III response Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 1 x p = 0.00, p = 0.00, p = 0.01, p = 0.00, p = 0.00 p = 0.00, p = 0.76, p = 0.00, p = 0.86, p = 0.00 p = 0.00, p = 1, p = 0.70, p = 0.11, p = 0.00 Cluster 2 x X p = 0.00, p = 0.00, p = 0.99, p = 0.00, p = 0.98 p = 0.31, p = 0.00, p = 0.05, p = 0.00, p = 0.01 Cluster 3 x X x p = 0.09, p = 0.68, p = 0.02, p = 0.00, p = 0.01 Cluster 4 x X x x Average alpha/low-beta and SEG effect size per hemisphere, stimulation amplitude (in mA), levodopa dose (in mg), MDS-UPDRS-III response (in points reduction) were pair-wise compared between clusters post-hoc using Tukey’s Honestly Significant Difference after ANOVA analyses revealed significant differences between clusters. The Tukey’s HSD allowed for multiple comparisons while controlling the probability of making one or more Type I errors. To summarize, alpha/low-beta physiomarkers in cluster 1 were associated with a relatively strong clinical response to levodopa despite active DBS. SEG physiomarkers in cluster 2 were associated with relatively small dose of medication and a small clinical response but a relatively high stimulation amplitude. Alpha/low-beta physiomarkers in cluster 3 were associated with a high levodopa dose but without a strong clinical effect (12). In cluster 4, hemispheres either contained an alpha/low-beta physiomarker or an SEG physiomarker without a clear predominance of the two. Therefore, this cluster did not reveal a distinct clinical profile relating to either alpha/low-beta or SEG physiomarkers. It is still unclear if some of these hemispheres could show an SEG physiomarker if stimulation amplitude increased, as entrainment has been shown to be a nonlinear process with patient-specific thresholds (29). Full spectrum response to levodopa (2-100 Hz) differs according to physiomarker frequency To further investigate the specificity of levodopa-induced spectral changes for the different physiomarker frequencies, the power values from the full spectrum (2-100 Hz) for all segments per hemisphere grouped by the physiomarker frequency band (theta, alpha/low-beta, high beta, low gamma, SEG or high gamma physiomarkers) were compared between OFF- and ON-MED with permutation testing (Supplementary Fig. 5). This figure shows that hemispheres with a physiomarker in the theta, high-beta, low-gamma and high gamma band, displayed broad spectral power changes after levodopa intake without clear modulation of low-beta or SEG. By contrast, no clear OFF-related peak in the beta band was observed for hemispheres with an SEG physiomarker. In the hemispheres with a physiomarker in either the alpha/low-beta or SEG band, a clear peak was seen centering around 62,5 Hz in the MED ON state, which reduced in the MED OFF state (29). In summary, depending on physiomarker frequency, distinct levodopa-induced changes were observed across the power spectrum suggesting different neurophysiological responses. Discussion In this study, we investigated the consistency of levodopa-induced changes in STN LFP power. A data-driven method was used to identify a 1-Hz frequency band in the STN LFP spectrum per hemisphere with the largest effect size when transitioning from the OFF to ON levodopa state during active DBS. We determined these ‘physiomarkers’ in three different states (30–60 seconds rest, randomly picked 30–60 seconds movement, and the full (3–7 minutes) MDS-UPDRS-III scoring) to investigate the effect of movement on physiomarker frequency, as beta oscillations have been shown to shift frequency with active movement ( 8 , 29 ), and identify if longer periods of data resulted in more stable levodopa-related spectral changes (Fig. 2 , Table 2 , Supplementary Figure S2-S4). The MDS-UPDRS-III score represents common daily movements and was used to assess the effect of levodopa and DBS in our center and, thus, representative of remaining treatment fluctuations that may be treated by aDBS. The most consistent set of physiomarkers when observing all frequency bands (i.e. consistent effect directions in the theta, alpha, low beta, high beta, low gamma, SEG and high gamma band) was extracted from LFPs recorded during the full MDS-UPDRS-III scoring (Table 2 ). Although across-hemisphere analyses did not reveal significant power changes across the broad frequency spectrum when comparing the OFF- to the ON levodopa state (Supplementary Figure S1 ), the within-hemisphere analyses revealed a personalised, statistically significant, levodopa-related physiomarker in every hemisphere. Detected physiomarkers most commonly had a frequency in the alpha/low-beta band (9–20 Hz, 41% of hemispheres), primarily showing suppression of power in the ON-dopaminergic compared to OFF-dopaminergic state (35% of hemispheres). The second most common physiomarker frequency band was SEG (20% of hemispheres), primarily showing an increase in power towards the ON-dopaminergic state compared with the OFF-dopaminergic state (19% of hemispheres). A slightly smaller percentage of hemispheres (19%) showed consistent increases in broadband high beta to low gamma power, while a minority of hemispheres had a physiomarker in the theta band (8%). In 12% of hemispheres the physiomarker frequency was found in the high gamma band but with conflicting effect directions. Next, we performed a k-means clustering analysis to identify clinical characteristics associated with the effect size of levodopa-induced spectral changes in the alpha/low-beta and SEG band. Four clusters were detected, for which two clusters linked the occurrence of alpha/low-beta physiomarkers with a relatively strong clinical response to levodopa or a high levodopa dose but weak clinical response, respectively. A third cluster linked the occurrence of SEG physiomarkers with a relatively small levodopa dose and clinical response but high stimulation amplitudes. However, the majority of hemispheres were assigned to a fourth cluster that contained physiomarkers in both the alpha/low-beta and SEG band, precluding any associations for these hemispheres with clinical variables. An important consideration here is that entrainment leading to an SEG physiomarker may depend on patient-specific stimulation thresholds ( 29 ). Therefore, some of these hemispheres may not have reached the appropriate stimulation amplitude for entrainment ( 29 ). Finally, while there was some SEG activity in hemispheres that showed an alpha/low-beta physiomarker, alpha/low-beta activity was negligible in hemispheres that showed an SEG physiomarker (Supplementary Figure S5). This corresponds with literature ( 9 , 14 ), showing minimal contribution of beta power to distinguishing the OFF- from the ON-dopaminergic state when SEG is present. Alpha/low-beta physiomarkers can be used in the currently available aDBS system ( 30 ), while SEG physiomarkers cannot. Therefore, our results support the use of alpha/low-beta based aDBS based upon levodopa fluctuations in a sizable number of patients (35% of our sample). This is in contrast with several, smaller, reports that found beta to be a poor physiomarker for aDBS ( 14 , 29 , 31 – 33 ). In the resting state, there was less consistency across hemispheres of physiomarker effect directions in the high-beta and low-gamma band compared to the movement state, possibly related to aperiodic activity (Supplementary Figure S5) ( 13 , 34 – 37 ). Typically, in this frequency band no distinct spectral peaks are observed but the power follows a characteristic 1/f-like pattern that is associated with aperiodic neural activity. The direction of the medication effect showed an increase in power after levodopa intake, which was also found in the paper by Gerster et al. ( 37 ). Furthermore, we found during the full MDS-UPDRS-III scoring that theta frequencies increased with power after levodopa intake (8% of hemispheres), similar to high-beta and low-gamma oscillations. The only commercially available aDBS system (‘Percept TM ’) currently does not correct for aperiodic activity when adapting the stimulation based on the power of a physiomarker. However, using long periods (minutes) for measuring physiomarker activity might result in more reliable estimates of the medication state (corresponding with our results regarding the consistency of physiomarker distributions in different states, Table 2 ). This may explain the significant improvement of ON-time duration without troublesome dyskinesia in the slower, dual-threshold (DT) aDBS compared to continuous DBS (cDBS) that was observed in the ADAPT PD trial while in the faster, single-threshold (ST) aDBS this was not observed ( 33 ). The majority of physiomarkers were found in the alpha/low-beta or SEG band. The occurrence of similar effect directions in the alpha and low-beta band may point towards similar neural processes, possibly explained by the frequently observed shift in beta frequencies when stimulation is turned on ( 8 , 38 ). Therefore, determining the physiomarker during stimulation, but also during movement, is key for picking the right frequency band used for aDBS ( 14 ). To generalize these findings to the clinical application of aDBS, longitudinal recordings are necessary to investigate the consistency of the personalized physiomarker to treatment alterations and symptom burden. For low beta and SEG power, this has already been established in previous work ( 12 , 14 ). However, for theta or broadband high beta to low gamma power this has not been investigated yet (Supplementary Fig. 5A, C, D). The contribution of theta, alpha and high beta in explaining the variance in patients' hemibody impairment OFF medication and the change in hemibody impairment following levodopa was reported previously, however, without active DBS and without showing the marked inter-individual differences in physiological responses to levodopa in PD that we show in this study ( 39 ). It is unknown if our results are applicable to GPi recordings, because to our knowledge, no clear modulation of beta power to levodopa has been shown to date. Furthermore, multiple factors may influence levodopa-related LFP physiomarker fluctuations, especially sleep, movement and stress can affect beta power ( 13 , 40 , 41 ). These effects and the physiomarkers that respond to them should be further investigated to determine the optimal feedback signal for aDBS. Although this study did not examine symptom specificity in relation to changes in specific frequency bands, it is plausible that a personalized model of STN activity—across different clinical states and accounting for both rest and movement—could more accurately inform adaptive DBS compared to current approaches that rely solely on band-based frequency features ( 42 – 45 ). Recent machine learning studies already showed marked improvement in classification of the OFF- and ON-dopaminergic state and motor performance when including the full frequency spectrum and burst time-dynamics ( 39 , 43 , 46 ). Future research could consider deep learning techniques that are already being applied for longitudinal Parkinson’s disease severity monitoring using other types of data, such as kinematics ( 47 ). Limitations Our data-driven method for determining personalized physiomarkers of the dopaminergic state during active DBS revealed that the neurophysiological response to levodopa varies substantially between individuals. However, there are some caveats arising from the methods we used. To start, the short segment of data (2 seconds) used for the calculation of the power spectral densities (PSDs) may have led to spectral leakage and insufficient frequency resolution for accurately characterizing narrowband oscillatory phenomena, such as SEG, and peak frequency estimation. Shorter time segments, however, provide greater temporal specificity for capturing transient burst dynamics that respond to levodopa, particularly in the beta band. Currently the only commercially approved Percept ™ system for aDBS uses a similar frequency resolution (0.98 Hz) and PSD overlap (80%) for the dual threshold algorithm and sensing durations for altering stimulation amplitude between 1 and 2 seconds (‘Onset duration’), corresponding largely with the PSD calculation in this study ( 30 ). Also, the configuration of the DBS stimulation may have been suboptimal in some patients despite completing cDBS titration, possibly causing the levodopa-related spectral fluctuations to be different and less disease-specific than those from the optimal contact used for stimulation. The heterogeneity in the duration of the recordings across patients during full MDS-UPDRS-III testing may have affected the accuracy of some physiomarkers, since this duration varied from about 3 to 7 minutes. Nevertheless, several minutes of data would still provide more than 100 PSDs per hemisphere for the Cohen’s d effect size computation. Besides technical factors, the variability in symptom profiles across patients likely influenced the detection of a physiomarker in the alpha/low-beta or SEG bands, which was not corrected or stratified for. For example, patients with tremor-predominant phenotypes tend to show a weaker association between LFP frequency bands and dopaminergic state compared to those with bradykinesia- or rigidity-predominant symptoms ( 48 ). Conclusion This study is the largest to date (n = 100 hemispheres) to investigate the frequency-specificity of levodopa-induced changes in STN spectral power at an individual level. We found that for 35% of hemispheres, the largest change in 1-Hz bin power after levodopa intake, referred to as “physiomarker”, was a reduction in power in the alpha/low-beta band, while 19% had an increase in SEG. Remaining physiomarkers were found in other frequency bands. Therefore, aDBS that can be tailored according to the individual neurophysiological profile beyond solely beta frequencies may improve aDBS outcome. The majority (56%) of physiomarkers exhibited a MED ON > MED OFF effect after levodopa intake, which is incompatible with current commercially-approved aDBS software. The neurophysiological signature identifying the dopaminergic state during active stimulation varies substantially between individuals and predicting the right physiomarker with just clinical characteristics may not be possible for all patients. This opens the door to machine-learning–based identification of disease states for aDBS and calls for commercial software that enables stimulation control based on multiple, and potentially inverse, LFP frequency features. Declarations Data availability Data is not available for public sharing due to ethical restrictions. Code availability The underlying code for this study is not publicly available but may be made available to qualified researchers on reasonable request from the corresponding author. Author Contributions MGJdN contributed to the design, execution, data acquisition, analysis, writing and editing of the final manuscript. BK, DH and MJS contributed to the data acquisition and editing of the final manuscript. RMAdB and RS contributed to the design and editing of the final manuscript. CRO, BCMvW, AWB and MB contributed to the design, writing and editing of the final manuscript. Funding This work was supported by the Amsterdam UMC TKI-PPP grant (2021 & 2023 call, Innovation Exchange Amsterdam). The funder played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript. Competing interests RS is consultant in the educational programs of Medtronic and Boston Scientific MB additionally received research funding related to the manuscript from Medtronic (2023, 2024). MGJdN, MJS, BK, DH, AWB, BCMvW, RMAdB have no conflicts of interests to declare in relation to the manuscript. References Limousin, P., Foltynie, T. Long-term outcomes of deep brain stimulation in Parkinson disease. Nature Reviews Neurology. 2019;15(4):234–42. Mansouri, A. et al. Deep brain stimulation for Parkinson’s disease: meta-analysis of results of randomized trials at varying lengths of follow-up. Journal of Neurosurgery JNS. 2018;128(4):1199–213. Deuschl, G. et al. European Academy of Neurology/Movement Disorder Society - European Section guideline on the treatment of Parkinson's disease: I. Invasive therapies. Eur J Neurol. 2022;29(9):2580–95. Hauser, R.A., McDermott, M.P., Messing, S. Factors associated with the development of motor fluctuations and dyskinesias in Parkinson disease. Arch Neurol. 2006;63(12):1756–60. Weaver, F.M. et al. Bilateral deep brain stimulation vs best medical therapy for patients with advanced Parkinson disease: a randomized controlled trial. Jama. 2009;301(1):63–73. Follett, K.A. et al. Pallidal versus subthalamic deep-brain stimulation for Parkinson's disease. N Engl J Med. 2010;362(22):2077–91. Odekerken, V.J. et al. Subthalamic nucleus versus globus pallidus bilateral deep brain stimulation for advanced Parkinson's disease (NSTAPS study): a randomised controlled trial. Lancet Neurol. 2013;12(1):37–44. Mathiopoulou, V. et al. Modulation of subthalamic beta oscillations by movement, dopamine, and deep brain stimulation in Parkinson’s disease. npj Parkinson's Disease. 2024;10(1):77. van Wijk, B.C.M., de Bie, R.M.A., Beudel, M. A systematic review of local field potential physiomarkers in Parkinson's disease: from clinical correlations to adaptive deep brain stimulation algorithms. J Neurol. 2023;270(2):1162–77. Chen, Y. et al. Neuromodulation effects of deep brain stimulation on beta rhythm: A longitudinal local field potential study. Brain Stimul. 2020;13(6):1784–92. Lofredi, R. et al. Subthalamic beta bursts correlate with dopamine-dependent motor symptoms in 106 Parkinson’s patients. npj Parkinson's Disease. 2023;9(1):2. de Neeling, M.G.J. et al. Beta Power Response After Levodopa Intake in Parkinson's Disease Patients With Chronic Sensing-Enabled Deep Brain Stimulation. Annals of Neurology. 2025;n/a(n/a). Belova, E.M., Semenova, U., Gamaleya, A.A., Tomskiy, A.A., Sedov, A. Voluntary movements cause beta oscillations increase and broadband slope decrease in the subthalamic nucleus of parkinsonian patients. Eur J Neurosci. 2021;53(7):2205–13. Oehrn, C.R. et al. Chronic adaptive deep brain stimulation versus conventional stimulation in Parkinson’s disease: a blinded randomized feasibility trial. Nature Medicine. 2024;30(11):3345–56. Sermon, J.J. et al. Sub-harmonic entrainment of cortical gamma oscillations to deep brain stimulation in Parkinson's disease: Model based predictions and validation in three human subjects. Brain Stimul. 2023;16(5):1412–24. Wiest, C., et al. Finely-tuned gamma oscillations: Spectral characteristics and links to dyskinesia. Exp Neurol. 2022;351:113999. von Elm, E., Altman, D.G., Egger, M., Pocock, S.J., Gøtzsche, P.C., Vandenbroucke, J.P. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007;370(9596):1453–7. Stam, M.J. et al. AI-DBS study: protocol for a longitudinal prospective observational cohort study of patients with Parkinson's disease for the development of neuronal fingerprints using artificial intelligence. BMJ Open. 2025;15(5):e091563. Holewijn, R.A. et al. General Anesthesia vs Local Anesthesia in Microelectrode Recording-Guided Deep-Brain Stimulation for Parkinson Disease: The GALAXY Randomized Clinical Trial. JAMA Neurol. 2021;78(10):1212–9. Goyal, A. et al. The development of an implantable deep brain stimulation device with simultaneous chronic electrophysiological recording and stimulation in humans. Biosens Bioelectron. 2021;176:112888. Oostenveld, R., Fries, P., Maris, E., Schoffelen, J.M. FieldTrip: Open Source Software for Advanced Analysis of MEG, EEG, and Invasive Electrophysiological Data. Computational Intelligence and Neuroscience. 2011;2011(1):156869. Stam, M.J. et al. A comparison of methods to suppress electrocardiographic artifacts in local field potential recordings. Clinical Neurophysiology. 2023;146:147–61. Maris, E., Oostenveld, R. Nonparametric statistical testing of EEG- and MEG-data. Journal of Neuroscience Methods. 2007;164(1):177–90. Gerber, E.M. permutest Retrieved November 11, 2024 ed: ( https://www.mathworks.com/matlabcentral/fileexchange/71737-permutest ), MATLAB Central File Exchange.; 2024. Lakens, D. Calculating and reporting effect sizes to facilitate cumulative science: a practical primer for t-tests and ANOVAs. Front Psychol. 2013;4:863. Schinkel, M., Appelman, B., Butler, J., Schuurman, A., Wiersinga, W.J. Association of clinical sub-phenotypes and clinical deterioration in COVID-19: further cluster analyses. Intensive Care Med. 2021;47(4):482–4. Rousseeuw, P.J. Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. Journal of Computational and Applied Mathematics. 1987;20:53–65. Ding, C., He, X., editors. K-means clustering via principal component analysis. Proceedings of the twenty-first international conference on Machine learning; 2004. Mathiopoulou, V. et al. Gamma entrainment induced by deep brain stimulation as a biomarker for motor improvement with neuromodulation. Nature Communications. 2025;16(1):2956. Stanslaski, S. et al. Sensing data and methodology from the Adaptive DBS Algorithm for Personalized Therapy in Parkinson’s Disease (ADAPT-PD) clinical trial. npj Parkinson's Disease. 2024;10(1):174. Colombo, A. et al. Finely Tuned γ Tracks Medication Cycles in Parkinson's Disease: An Ambulatory Brain-Sense Study. Mov Disord. 2025;40(5):881–95. Gerster, M. et al. Beyond beta rhythms: Aperiodic broadband power reflects Parkinson’s disease severity–a multicenter study. bioRxiv. 2025:2025.03.11.642600. Bronte-Stewart H.M. et al. Long-Term Personalized Adaptive Deep Brain Stimulation in Parkinson Disease: A Nonrandomized Clinical Trial. JAMA Neurology. 2025. Wiest, C. et al. The aperiodic exponent of subthalamic field potentials reflects excitation/inhibition balance in Parkinsonism. Elife. 2023;12. Stam, M.J. et al. Long-term consistency of aperiodic and periodic physiomarkers in subthalamic local field potentials in Parkinson’s disease. npj Parkinson's Disease. 2025;11(1):204. Kim, J. et al. Dopamine depletion can be predicted by the aperiodic component of subthalamic local field potentials. Neurobiology of Disease. 2022;168:105692. Gerster, M. et al. Beyond beta rhythms: subthalamic aperiodic broadband power scales with Parkinson's disease severity-a cross-sectional multicentre study. EBioMedicine. 2025;122:105988. Feldmann, L.K. et al. Toward therapeutic electrophysiology: beta-band suppression as a biomarker in chronic local field potential recordings. npj Parkinson's Disease. 2022;8(1):44. Khawaldeh, S. et al. Balance between competing spectral states in subthalamic nucleus is linked to motor impairment in Parkinson's disease. Brain. 2022;145(1):237–50. Baumgartner, A.J., Hirt, L., Amara, A.W., Kern, D.S., Thompson, J.A. Diurnal fluctuations of subthalamic nucleus local field potentials follow naturalistic sleep-wake behavior in Parkinson’s disease. Sleep. 2025;48(6):zsaf005. Feldmann LK, Lofredi R, Al-Fatly B, Busch JL, Mathiopoulou V, Roediger J, et al. Christmas-Related Reduction in Beta Activity in Parkinson's Disease. Mov Disord. 2023;38(4):692–7. Merk, T. et al. Machine learning based brain signal decoding for intelligent adaptive deep brain stimulation. Experimental Neurology. 2022;351:113993. Sand, D. et al. Machine learning-based personalized subthalamic biomarkers predict ON-OFF levodopa states in Parkinson patients. J Neural Eng. 2021;18(4). Peterson, V. et al. Movement decoding using spatio-spectral features of cortical and subcortical local field potentials. Exp Neurol. 2023;359:114261. Toth, R. et al. DyNeuMo Mk-2: An Investigational Circadian-Locked Neuromodulator with Responsive Stimulation for Applied Chronobiology. Conf Proc IEEE Int Conf Syst Man Cybern. 2020;2020:3433–40. Castaño-Candamil, S. et al. Identifying controllable cortical neural markers with machine learning for adaptive deep brain stimulation in Parkinson's disease. Neuroimage Clin. 2020;28:102376. Franco, A. et al. The Role of Deep Learning and Gait Analysis in Parkinson’s Disease: A Systematic Review. Sensors. 2024;24(18):5957. Beudel, M. et al. Oscillatory Beta Power Correlates With Akinesia-Rigidity in the Parkinsonian Subthalamic Nucleus. Mov Disord. 2017;32(1):174–5. Additional Declarations Competing interest reported. RS is consultant in the educational programs of Medtronic and Boston Scientific MB additionally received research funding related to the manuscript from Medtronic (2023, 2024). MGJdN, MJS, BK, DH, AWB, BCMvW, RMAdB have no conflicts of interests to declare in relation to the manuscript. Supplementary Files Supplementary.pdf Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 20 Apr, 2026 Reviews received at journal 20 Apr, 2026 Reviews received at journal 11 Apr, 2026 Reviewers agreed at journal 07 Apr, 2026 Reviews received at journal 07 Apr, 2026 Reviewers agreed at journal 03 Apr, 2026 Reviews received at journal 02 Apr, 2026 Reviewers agreed at journal 01 Apr, 2026 Reviewers agreed at journal 31 Mar, 2026 Reviewers invited by journal 31 Mar, 2026 Editor assigned by journal 26 Mar, 2026 Submission checks completed at journal 26 Mar, 2026 First submitted to journal 22 Mar, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9190790","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":617064701,"identity":"808db572-f157-4e2d-bef6-4cc54a91e465","order_by":0,"name":"Martijn G.J. de Neeling","email":"","orcid":"","institution":"Amsterdam University Medical Centers","correspondingAuthor":false,"prefix":"","firstName":"Martijn","middleName":"G.J.","lastName":"de Neeling","suffix":""},{"id":617064702,"identity":"46b6067e-5e33-46c6-b53c-ea3f85678dee","order_by":1,"name":"Carina R. Oehrn","email":"","orcid":"","institution":"University of California, Davis","correspondingAuthor":false,"prefix":"","firstName":"Carina","middleName":"R.","lastName":"Oehrn","suffix":""},{"id":617064703,"identity":"1176f236-b810-4351-8d05-50db29b4e60c","order_by":2,"name":"Mariëlle J. Stam","email":"","orcid":"","institution":"Amsterdam University Medical Centers","correspondingAuthor":false,"prefix":"","firstName":"Mariëlle","middleName":"J.","lastName":"Stam","suffix":""},{"id":617064704,"identity":"0e2b3ba0-f442-4f56-8b30-c2d1c484440b","order_by":3,"name":"Bart J. Keulen","email":"","orcid":"","institution":"Amsterdam University Medical Centers","correspondingAuthor":false,"prefix":"","firstName":"Bart","middleName":"J.","lastName":"Keulen","suffix":""},{"id":617064705,"identity":"f34f20d9-64da-4df4-9f6f-874ca9594bad","order_by":4,"name":"Deborah Hubers","email":"","orcid":"","institution":"Amsterdam University Medical Centers","correspondingAuthor":false,"prefix":"","firstName":"Deborah","middleName":"","lastName":"Hubers","suffix":""},{"id":617064706,"identity":"6396863c-0bd9-4001-b2db-7a28c7d7fd11","order_by":5,"name":"Rob M.A. Bie","email":"","orcid":"","institution":"Amsterdam University Medical Centers","correspondingAuthor":false,"prefix":"","firstName":"Rob","middleName":"M.A.","lastName":"Bie","suffix":""},{"id":617064707,"identity":"157e2270-8d9a-490e-af67-f2ed0effe5a7","order_by":6,"name":"Bernadette C.M. Wijk","email":"","orcid":"","institution":"Vrije Universiteit Amsterdam","correspondingAuthor":false,"prefix":"","firstName":"Bernadette","middleName":"C.M.","lastName":"Wijk","suffix":""},{"id":617064708,"identity":"023df6bc-d63d-4afe-b7cf-23f89dfaf1b2","order_by":7,"name":"Rick Schuurman","email":"","orcid":"","institution":"Amsterdam University Medical Centers","correspondingAuthor":false,"prefix":"","firstName":"Rick","middleName":"","lastName":"Schuurman","suffix":""},{"id":617064709,"identity":"36b285d5-b96a-4611-9acf-77d49928c725","order_by":8,"name":"Arthur W.G. Buijink","email":"","orcid":"","institution":"Amsterdam University Medical Centers","correspondingAuthor":false,"prefix":"","firstName":"Arthur","middleName":"W.G.","lastName":"Buijink","suffix":""},{"id":617064710,"identity":"302c2950-3900-4beb-aac1-d5f86e89fad9","order_by":9,"name":"Martijn Beudel","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA9ElEQVRIiWNgGAWjYFACHiBmY5Dhh3IZG6AMGUJaeCRBSg8gaeEhqMXgALFadBt4Dz7mKbPhMT5+Ovnzh4p7sv3Tzphu+MFwB6cWswN8ycY859J4zM7kbpM4cKbYeMbtHLObPQzP8GjhMZPmbTvMY3YgdxvDwbaExAagltsMDIfxaTH/zdv2n8e4/+3mDwf/JSTOJ0KLGTNv2wEeA4ncDRIHGxISNxDUcpgvWXLOuWQeiRtvt0mcOZZgvPF2WtnNHgM8Wo73HvzwpsxOjr8/d/OHipoE2Xm3k7fd+FFxWA6XFgZm7MIGODWMglEwCkbBKCACAABHvVwsVDVzyAAAAABJRU5ErkJggg==","orcid":"","institution":"Amsterdam University Medical Centers","correspondingAuthor":true,"prefix":"","firstName":"Martijn","middleName":"","lastName":"Beudel","suffix":""}],"badges":[],"createdAt":"2026-03-22 11:23:30","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9190790/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9190790/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":106300066,"identity":"62b2cb07-6c34-44a5-99a2-1c119d8fbb44","added_by":"auto","created_at":"2026-04-07 09:12:29","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":5431271,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eOverview of analysis pipeline. A)\u003c/strong\u003e For each patient in this study, BrainSense\u003csup\u003eTM \u003c/sup\u003eStreaming was used to record LFPs from the stimulation contact point in 2 clinical states: during active DBS (“ON DBS”) without recent (\u0026gt;12 hours) intake of levodopa (“MED OFF”) and 1 hour after intake (“STIM ON MED”). In total, 30 to 60 seconds were recorded in the resting state and 3 to 7 minutes during clinical examination with the MDS-UPDRS-III scale. \u003cstrong\u003eB) \u003c/strong\u003eThe recorded LFP time series were segmented into 2 second epochs for which a PSD was calculated each. The median and IQR of all these PSDs (approximately 120 to 240) for one hemisphere during ON STIM MED OFF (left) and ON STIM ON MED (right) is shown as an example. \u003cstrong\u003eC) \u003c/strong\u003eA cluster-based permutation test was performed on the effect size of medication (Cohen’s d) per 1 Hz frequency bin to identify the frequency bin with the largest effect size and within a significant cluster (‘the physiomarker’). An example for one hemisphere is shown. \u003cstrong\u003eD)\u003c/strong\u003e Visualization of the distribution of physiomarkers across hemispheres together with the direction of their effect. The occurrence of the physiomarkers per frequency band (theta, alpha, low beta, high beta, low gamma, SEG and high gamma) was statistically tested per state (30-60 seconds rest, 30-60 seconds movement, and full MDS-UPDRS scoring). \u003cstrong\u003eE) \u003c/strong\u003eConsistency of all physiomarker effect directions across hemispheres was assessed by calculating the mean sign of all effect directions (MED ON\u0026gt;MED OFF versus MED OFF\u0026gt;MED ON) across all physiomarkers within each frequency band. A two-tailed binomial test assessed whether the proportion of effect directions differed from 0. Adjacent frequency bands with a significant effect direction in the same direction (alpha and low beta) were collapsed for subsequent clustering analysis. \u003cstrong\u003eF) \u003c/strong\u003eK-means clustering was used to identify clinical characteristics (i.e., disease duration, levodopa dose, stimulation amplitude, MDS-UPDRS-III scores) associated with the occurrence and magnitude of physiomarkers in the alpha/low-beta and SEG bands.\u003c/p\u003e","description":"","filename":"figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-9190790/v1/2cba91d93e49f2d17866e064.png"},{"id":106300001,"identity":"4fdf6a15-b1e2-4918-bcc4-ec53089c7286","added_by":"auto","created_at":"2026-04-07 09:12:08","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1628289,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDistribution of physiomarkers indicating hemisphere-specific frequencies with the largest effect of medication on LFP power. \u003c/strong\u003eShown are the percentage of hemispheres (n = 100) for which the physiomarker was detected within the indicated frequency band. The direction of the effect (MED ON\u0026gt;MED OFF, MED OFF\u0026gt;MED ON) is\u003cstrong\u003e \u003c/strong\u003ecolour-coded. \u003cstrong\u003e(A)\u003c/strong\u003e Distribution of physiomarkers during the resting state (30 seconds – 1 minute). \u003cstrong\u003e(B)\u003c/strong\u003eDistribution of physiomarkers during movement (30 seconds – 1 minute, randomly sampled during MDS-UPDRS-III scoring to match the duration of the resting state).\u003cstrong\u003e (C)\u003c/strong\u003e Distribution of physiomarkers during the full MDS-UPDRS-III scoring (3 – 7 minutes). Compared to panel A and B, more consistent effect size directions were found across frequency bands.\u003c/p\u003e","description":"","filename":"figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-9190790/v1/3f8057511072a4625d0a1204.png"},{"id":106300061,"identity":"e3e2175d-0cbd-4357-a197-692567d94819","added_by":"auto","created_at":"2026-04-07 09:12:27","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":2992393,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ek-means clustering analysis linking physiomarker effect size to clinical characteristics.\u003c/strong\u003e \u003cstrong\u003eA)\u003c/strong\u003e Visualization of the k-means clusters in principal component analysis (PCA) space. The first 3 principal components accounted for 74% of the variance of the data. Cluster separation is illustrated by the boundary outlines. \u003cstrong\u003eB) \u003c/strong\u003eAverage alpha/low-beta effect size per hemisphere differed significantly between clusters (ANOVA p-value is indicated). Pair-wise post-hoc tests revealed a lower average effect size per hemisphere in cluster 1 compared to all other clusters and a lower average effect size per hemisphere in cluster 3 compared to cluster 2. \u003cstrong\u003eC) \u003c/strong\u003eAverage SEG effect size per hemisphere differed significantly between clusters (ANOVA p-value is indicated). Pair-wise post-hoc tests revealed a higher average effect size per hemisphere in cluster 2 compared to all other clusters.\u003cstrong\u003e D) \u003c/strong\u003eThe first three clusters showed a clear physiomarker composition: hemispheres with a physiomarker primarily in the alpha/low-beta band in cluster 1 and 3, hemispheres with a physiomarker in the SEG band in cluster 2, while cluster 4 contained a mixture of hemispheres with a physiomarker in either the alpha/low-beta or SEG band. \u003cstrong\u003eE) \u003c/strong\u003eMedication dose significantly differed between clusters (ANOVA p-value is indicated). Pair-wise post-hoc tests revealed a higher median dose in cluster 1 and 3 compared to cluster 2 and 4. \u003cstrong\u003eF)\u003c/strong\u003eStimulation amperage significantly differed between clusters (ANOVA p-value is indicated). Pair-wise post-hoc tests revealed a lower stimulation amplitude in cluster 1 compared to cluster 2 and 3 and a trend towards a higher stimulation amplitude in cluster 2 and 3 compared to cluster 4. \u003cstrong\u003eG)\u003c/strong\u003e MDS-UPDRS-III hemibody responses to levodopa significantly differed between the clusters (ANOVA p-value is indicated). Pair-wise post-hoc tests revealed lower MDS-UPDRS-III hemibody score reductions in response to levodopa in cluster 2 and 3 compared to cluster 1 and 4.\u003c/p\u003e","description":"","filename":"figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-9190790/v1/288d46cd77a22a58c68c8041.png"},{"id":106300499,"identity":"1d9b2e90-b118-4def-a95d-8fe5f868957b","added_by":"auto","created_at":"2026-04-07 09:13:57","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":11342307,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9190790/v1/b79f6ddd-ad48-4849-aa0b-d3ed43ccff8d.pdf"},{"id":106300065,"identity":"bfaafc17-6af3-4026-9f02-9b81398b724a","added_by":"auto","created_at":"2026-04-07 09:12:29","extension":"pdf","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":4508057,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementary.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9190790/v1/1ebd764ea9567b2d2c7b72db.pdf"}],"financialInterests":"Competing interest reported. RS is consultant in the educational programs of Medtronic and Boston Scientific\nMB additionally received research funding related to the manuscript from Medtronic (2023, 2024). \nMGJdN, MJS, BK, DH, AWB, BCMvW, RMAdB have no conflicts of interests to declare in relation to the manuscript.","formattedTitle":"Levodopa-related physiomarkers during deep brain stimulation in Parkinson’s disease","fulltext":[{"header":"Introduction","content":"\u003cp\u003eDeep brain stimulation (DBS) of the subthalamic nucleus (STN) is an effective treatment for advanced Parkinson\u0026rsquo;s disease (PD)(\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e). It is particularly beneficial for managing levodopa-related motor fluctuations, such as medication-induced dyskinesia and bradykinesia related to medication wearing off (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). Yet, patients receiving DBS may still experience remaining motor fluctuations, often due to stimulation-induced side effects that prevent further increases in stimulation amplitude (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan additionalcitationids=\"CR6\" citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e). Therefore, adapting stimulation amplitude according to the disease state (e.g. higher stimulation when medication is wearing off) could improve the efficacy of DBS. Adaptive DBS (aDBS) automatically adjusts the stimulation amplitude based upon specific neurophysiological markers of symptom severity, referred to as physiomarkers (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e). To date, the most established physiomarker is the spectral power within the beta (\u0026plusmn;\u0026thinsp;13\u0026ndash;30 Hz) frequency band of STN local field potentials (LFPs) that captures the occurrence and strength of beta oscillations. This is based on several studies showing that, in the absence of stimulation and medication (STIM OFF, MED OFF), beta power is positively correlated with the severity of (contralateral) bradykinesia and rigidity (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e) and reduced by therapeutic interventions such as medication and DBS (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan additionalcitationids=\"CR11\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). Nevertheless, the magnitude of beta power explains only\u0026thinsp;~\u0026thinsp;17% of variability in symptom severity across patients (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e). Furthermore, beta oscillations are co-modulated by voluntary movement, which complicates use of this physiomarker for aDBS (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). In line, a recent study (n\u0026thinsp;=\u0026thinsp;4) found that STN beta power measured in ambulatory settings was only predictive of motor symptoms in one of six hemispheres (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eRecent studies have pointed out \u003cem\u003estimulation-entrained gamma\u003c/em\u003e (SEG) oscillations (typically at 62.5 Hz with 125 Hz stimulation frequency, a subharmonic) as an alternative physiomarker of symptom severity and medication state (\u003cspan additionalcitationids=\"CR15\" citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e). A first aDBS feasibility study in the at-home environment demonstrated that SEG recorded in the STN and sensorimotor cortex using electrocorticography (ECoG), showed sharp increases after levodopa intake and decreases when levodopa is wearing off and was more informative of the patient\u0026rsquo;s medication state than beta power (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). Although this pilot study (n\u0026thinsp;=\u0026thinsp;4) suggested SEG may be preferable over the use of beta power as physiomarker for aDBS, it is not yet known how many patients exhibit SEG activity in the STN or which factors contribute to its occurrence. In order to identify the optimal physiomarker related to the dopaminergic state during stimulation in a larger cohort, we investigated the effect of medication intake on the spectral power of LFP signals in a large cohort of 52 patients (100 hemispheres) using a data-driven analysis without a-priori assumption of the physiomarker frequency (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). In addition, we used a k-means clustering method to explain the occurrence and magnitude of the two most commonly found physiomarkers by the clinical characteristics of the patients.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eParticipants\u003c/h2\u003e \u003cp\u003eThe STROBE criteria were used in the reporting of this study (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e). Patients with Parkinson\u0026rsquo;s disease and bilateral DBS surgery were included between November 2022 and November 2024 as part of the AI-DBS study (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e). SenSight directional leads (model 33005, Medtronic, Minneapolis, MN, US) were implanted in the subthalamic nucleus (STN) together with the Medtronic Percept\u003csup\u003e\u0026trade;\u003c/sup\u003e PC neurostimulator using previously described procedures followed by our center (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e) in the context of standard clinical care. This study was approved by the local ethics committee (NL80384.018.22) and was carried out in accordance with the Declaration of Helsinki. Informed written consent was received from all patients. Exclusion criteria included the lack of clinical data or the (full) LFP recording at the follow-up time point after continuous DBS (cDBS) optimization, use of a continuous intrajejunal or subcutaneous (fos)levodopa, and serious adverse events after DBS surgery.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eData acquisition\u003c/h3\u003e\n\u003cp\u003eThe data described in this study were collected during the 6-month follow-up visit after DBS implementation. Variability in motor symptom severity was assessed by scoring the Movement Disorders Society Unified Parkinson\u0026rsquo;s Disease Rating Scale part III (MDS-UPDRS-III) while sensing LFP time series with BrainSense\u003csup\u003e\u0026trade;\u003c/sup\u003e Streaming in two conditions: ON stimulation and without medication (\u0026ldquo;STIM ON MED OFF\u0026rdquo;) and ON stimulation and with medication (STIM ON MED ON) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The MED ON state was assessed 1 hour after intake of a suprathreshold (120%) dose of the morning levodopa equivalent daily dose (LEDD) with a minimum of 100 mg. Each patient started with the STIM ON MED OFF scoring. For each of the two conditions, we conducted a recording at rest for 1 minute and for 3 to 7 minutes while assessing the MDS-UPDRS-III score. LFPs were recorded using the surrounding contacts of the contact used for stimulation. The LFP signal was high-pass filtered at 1 Hz and low-pass filtered at 100 Hz, sampled at 250 Hz, and stored on the Percept\u003csup\u003e\u0026trade;\u003c/sup\u003e device. After the session, the data were retrieved, together with the stimulation amplitude, from the clinician programmer (\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e). The suprathreshold levodopa dose per patient was extracted from clinical records. We analyzed video recordings to determine the start- and end time of the minute resting state and the MDS-UPDRS-III scoring. Synchronization of video- and LFP data was performed by filming the onset of the recording prior to clinical examination.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eData processing\u003c/h3\u003e\n\u003cp\u003eWe used MATLAB (R2023b, The MathWorks, Inc., Natick,MA, USA) and the FieldTrip toolbox (version 1.0.2.0 (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e)) for our analyses. Patients with insufficient resting state data, defined by less than 30 seconds starting from the beginning of each resting state recording, were excluded from the analysis. We first utilised a single value decomposition method for suppressing ECG artifacts within the recordings (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e). Then, we computed the power spectral density (PSD) of non-interrupted time segments based on all available data per hemisphere using Welch\u0026rsquo;s method, a window size of 1 second and 95% overlap, resulting in a PSD over 1-100 Hz with 1-Hz frequency bins. These single PSD estimates per hemisphere served to compare the MED OFF and MED ON state on a group level using all data available (without separation of rest and movement data). For this, each PSD was normalized by dividing each power value by the maximum value per STN. Subsequently, the median power and IQR was determined across hemispheres per frequency bin.\u003c/p\u003e \u003cp\u003eWe then repeated the PSD computation after dividing the available time series into non-overlapping epochs of 2 seconds duration, hence resulting in multiple PSD estimates per hemisphere. For each hemisphere, outlying power values exceeding five standard deviations were replaced by the median value across PSDs per frequency bin. The set of PSD estimates per hemisphere served to determine a personalized, data-driven physiomarker of the medication state. For this, we used unnormalized spectra and focused on the 2-100 Hz frequency range. These PSDs were computed separately for the resting state (30 seconds\u0026thinsp;\u0026minus;\u0026thinsp;1 minute), movement state with a matched duration to the resting state, and the full duration of the MDS-UPDRS-III scoring (3\u0026ndash;7 minutes of data). The duration-matched movement state was randomly selected from the total MDS-UPDRS-III scoring segment to make sure that detected physiomarkers were not based upon one specific movement task.\u003c/p\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e \u003ch2\u003ePhysiomarker detection\u003c/h2\u003e \u003cp\u003eIn order to determine personalised physiomarkers of the medication state, we used cluster-based non-parametric permutation tests (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e) as implemented in the \u003cem\u003epermutest.m\u003c/em\u003e (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e) function (10.000 permutations, t-statistic, two-sided paired T-test, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) to statistically compare OFF- and ON medication state PSDs for each hemisphere separately. We used the same permutation test for determining spectral changes from OFF- to ON MED across hemispheres (one PSD per hemisphere) to investigate the difference between within- and across-hemispheres results. We determined the magnitude of the statistical difference per hemisphere by calculating Cohen\u0026rsquo;s d effect size per 1-Hz bin and selected the frequency bin with the largest effect size within a statistically significant cluster as the \u0026ldquo;physiomarker\u0026rdquo;, similar to Oehrn et al. (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). This could result in multiple significant frequency bins per hemisphere in case multiple significant clusters were detected, however, only the largest effect size within the power spectrum from all significant clusters was selected as physiomarker. We regarded any effect size between \u0026minus;\u0026thinsp;0.2 and 0.2 too small and therefore non-significant, hence leaving the possibility of hemispheres with no detectable physiomarker (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). We then calculated the percentage of the total number of physiomarkers found (across all hemispheres) in the theta (4\u0026ndash;8 Hz), alpha (9\u0026ndash;12 Hz), low beta (13\u0026ndash;20 Hz), high beta (21\u0026ndash;30 Hz), low gamma (31\u0026ndash;59 Hz), SEG (60\u0026ndash;65 Hz with 125 Hz stimulation frequency (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e)) and high gamma (66\u0026ndash;100 Hz) band to investigate in which frequency band most physiomarkers were detected. The percentage of physiomarkers between states was compared per frequency band with a Kruskal-Wallis test and post-hoc Mann-Whitney U test.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eConsistency of physiomarkers across hemispheres\u003c/h2\u003e \u003cp\u003eNext, we investigated the predominant direction of effect per frequency band across all personalized physiomarkers to test for similarity in presumed underlying neuronal mechanisms. Combined neighboring frequency bands were used for further analyses if this resulted in a common significant effect direction opposed to classic frequency band definitions in literature that were determined without active stimulation (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e). Within-band effect directions were assessed by computing the mean sign of all effect directions across all physiomarkers within each band, i.e. discarding the magnitude of the effect size but only considering its directionality with +\u0026thinsp;1 indicating an increase in power after medication intake and \u0026minus;\u0026thinsp;1 a decrease. For each clinical state (rest, duration-matched movement, full MDS-UPDRS-III scoring), a two-tailed binomial test evaluated whether the observed proportion of MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF versus MED OFF\u0026thinsp;\u0026gt;\u0026thinsp;MED ON directions deviated significantly from an equal 50/50 distribution. Mean sign values closer to 0 were interpreted as greater directional conflict, whereas mean sign closer to -1/1 were taken to indicate more consistent effects across hemispheres. Only the mean sign values for which the binomial test indicated a significant deviation from a 50/50 distribution were considered a significant, common effect direction per frequency band.\u003c/p\u003e \u003cp\u003eTo assess which of these three datasets (resting state, matched-duration movement state or the full MDS-UPDRS-III scoring) most consistently identified medication state, we averaged the mean sign values across frequency bands, and selected the dataset with the highest value (full MDS-UPDRS-III scoring) for all subsequent analyses.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eClinical characteristics associated with physiomarker effect size and frequency\u003c/h3\u003e\n\u003cp\u003eWe performed a clustering analysis to discover distinguishing sets of clinical characteristics associated with physiomarkers across hemispheres (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e). Specifically, we tested for shared variance between several clinical variables and the magnitude of the physiomarker in the two frequency bands with most detected physiomarkers (i.e., alpha/low-beta and SEG). In hemispheres containing either the most common physiomarker (alpha/low-beta) or second most common physiomarker (SEG), for each hemisphere the mean effect size for all 1-Hz bins for each of these two frequency bands (alpha/low-beta and SEG) was combined with selected clinical features in a k-means clustering analysis to partition the data into groups based on their similarity. The clinical features included levodopa dose in mg, stimulation amplitude in milli-amperage, disease burden (MED OFF STIM OFF score), MDS-UPDRS-III response to levodopa during stimulation, and disease duration in years. The MDS-UPDRS-III response to levodopa during stimulation was defined as the STIM ON MED ON MDS-UPDRS-III hemibody score subtracted by the STIM ON MED OFF MDS-UPDRS-III hemibody score. The hemibody score was calculated by summing the rigidity, bradykinesia and tremor scores for either side of the body. The variables were pair-wise tested for collinearity (r\u0026thinsp;\u0026gt;\u0026thinsp;0.5), which was the case for disease burden, the MDS-UPDRS-III response to levodopa during stimulation, and disease duration. Only the MDS-UPDRS-III response to levodopa during stimulation was retained as variable to avoid collinearity of variables that may distort cluster formation and bias results. Prior to clustering, all features were standardized (z-scored) to ensure equal weighting across dimensions.\u003c/p\u003e \u003cp\u003eWe used the Silhouette score to measure how similar each observation is to its own cluster compared to other clusters across a range of cluster numbers (two to eight clusters, Supplementary Figure S6) (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e), which resulted in four clusters yielding the highest median score. The MATLAB function \u003cem\u003ekmeans.m\u003c/em\u003e was used with the k-means\u0026thinsp;+\u0026thinsp;+\u0026thinsp;initialization strategy to improve centroid placement and convergence. To reduce sensitivity to initialization and avoid convergence to local minima, the algorithm was repeated 100 times with different random initializations. The solution with the lowest within-cluster sum of squared distances was retained. To visualize the resulting clusters in a reduced-dimensional space, a principal component analysis (PCA) was applied to the feature matrix (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e). The first three principal components, which together explained more than 70% of the variance, were plotted in 3-D scatter plots. Cluster separation was further highlighted by overlaying boundary outlines around each cluster and color-coding individual points according to their cluster assignment. To statistically evaluate differences between clusters, a one-way analysis of variance (ANOVA, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) was performed on each feature. Post-hoc Tukey\u0026rsquo;s honestly significant difference (HSD, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) test was then applied to identify which cluster pairs significantly differed from each other. Box plots per clinical variable including all data points separated by their respective cluster were plotted for visualization. For each cluster, we counted the number of hemispheres with a physiomarker detected in the alpha/low-beta and SEG frequency bands, to obtain insights into the clinical variables associated with physiomarker frequency.\u003c/p\u003e\n\u003ch3\u003eDiscrimination of levodopa-related full spectral responses according to physiomarker frequency\u003c/h3\u003e\n\u003cp\u003eFinally, to investigate possible co-occurrence of multiple physiomarkers per hemisphere, we grouped the individual full power spectra (2-100 Hz) based on the frequency band of the detected physiomarker, and performed another cluster-based permutation test on the aggregated 2-second segmented PSD data. This data, without normalization, from all hemispheres per physiomarker frequency band (e.g. pooling all data from hemispheres that had an alpha/low-beta physiomarker (9\u0026ndash;20 Hz)) was compared between MED ON and MED OFF and the median and IQR across segments was plotted separately per medication state.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec12\"\u003e\n \u003ch2\u003eClinical characteristics\u003c/h2\u003e\n \u003cp\u003eAll data reported represent the median and interquartile range [IQR] unless otherwise specified. Data from fifty-two PD patients (31 male) with a total of 104 hemispheres (52 left, 52 right) were included in the study. Four hemispheres were excluded due to missing data, leaving 100 hemispheres (51 left, 49 right) for further analysis. Patients were 64 years old [59, 70] and had a disease duration of 10 [8, 14] years. During preoperative screening, the MDS-UPDRS-III MED OFF score was 43 [36, 53] and the MDS-UPDRS-III ON MED score was 18 [12, 25], showing an improvement after levodopa intake of 59% [51, 66%]. The MDS-UPDRS-III ON STIM MED OFF score after DBS titration (clinical follow up) was 23 [14, 30], showing a reduction of 49% [35, 50%] compared to the pre-surgical MDS-UPDRS-III score. The MDS-UPDRS-III ON STIM ON MED score at follow up was 19 [13, 23]. The levodopa equivalent daily dose (LEDD) during pre-operative screening was 1332 mg [935, 1674 mg] and 489 mg [300, 686 mg] at clinical follow up, showing a reduction of 60% [50, 75%]. Stimulation amplitude at follow up was 2.0 mA [1.6, 2.4 mA] and stimulation frequency was 125 Hz in all patients. A group analysis of PSDs across hemispheres (n\u0026thinsp;=\u0026thinsp;100, 2-100 Hz) did not show a statistically significant difference between the STIM ON MED OFF phase compared to the STIM ON MED ON phase (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, Supplementary Fig.\u0026nbsp;1). PSDs determined per hemisphere using the segmentation method (Fig.\u0026nbsp;1) resulted in a median of 183 (IQR [162, 232]) PSDs per hemisphere. Findings for the distribution of physiomarkers per dataset are summarized in Fig.\u0026nbsp;2.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\"\u003e\n \u003ch2\u003eDistribution of physiomarkers during the resting state\u003c/h2\u003e\n \u003cp\u003eTwo peaks were observed in the distribution of physiomarkers detected from the resting state LFP segments, one in the alpha/low-beta band and one in the SEG band (Fig.\u0026nbsp;2A; Table\u0026nbsp;1). For 13% of hemispheres the largest effect size of medication was within the alpha band, and in 18% of hemispheres in the low beta band. MED OFF\u0026thinsp;\u0026gt;\u0026thinsp;MED ON physiomarkers most commonly occurred at 14 Hz within the low beta band. For 11% of hemispheres the largest effect of medication was in the SEG band with a MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF effect peak at 62 Hz. For 45% of hemispheres, the largest effect size was not found in the alpha, low beta or SEG band, but in the theta (15%), high beta (13%), low gamma (17%) or high gamma (13%) band. In general, physiomarkers were spread across a broad frequency range with only clear effect size directions in the theta, alpha-beta and SEG band (Fig.\u0026nbsp;2A). The median stimulation amplitude was 1.9 mA [1.5, 2.5 mA] for hemispheres with physiomarkers in the alpha-beta band and 2.3 mA [1.7, 2.6 mA] for hemispheres with physiomarkers in the SEG band without a significant difference (MWU test, p\u0026thinsp;\u0026gt;\u0026thinsp;0.10).\u003c/p\u003e\n \u003cdiv\u003e\n \u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 1\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026ndash; Percentage of hemispheres with detected physiomarkers per frequency band.\u003c/strong\u003e Percentages are expressed compared to the total number of hemispheres in the analysis (n\u0026thinsp;=\u0026thinsp;100) and listed separately for the three different states (resting state, movement, and full MDS-UPDRS-III scoring). The direction of the medication effect on LFP power of the physiomarker is specified between brackets, with \u0026darr; illustrating a decrease in power when transitioning to the MED ON state and \u0026uarr; illustrating an increase in power when transitioning to the MED ON state. Low beta showed higher occurrence during full MDS-UPDRS-III scoring compared to the resting state (p\u0026thinsp;=\u0026thinsp;0.03, indicated with an asterisk), while SEG physiomarkers showed a trend towards more occurrence between these two states (p\u0026thinsp;=\u0026thinsp;0.08). A significant physiomarker was detected in all hemispheres (not significant\u0026thinsp;=\u0026thinsp;0%).\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eFrequency band\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eResting state\u003c/p\u003e\n \u003cp\u003e(30 sec \u0026ndash; 1 min)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003eMovement\u003c/p\u003e\n \u003cp\u003e(30 sec \u0026ndash; 1 min)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003eFull MDS-UPDRS-III scoring\u003c/p\u003e\n \u003cp\u003e(3 min \u0026ndash; 7 min)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eTheta (4\u0026ndash;8 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e15% (1% \u0026darr;, 14%\u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e9% (1% \u0026darr;, 8% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e8% (0% \u0026darr;, 8% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eAlpha (9\u0026ndash;12 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e13% (10% \u0026darr;, 3% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e8% (5% \u0026darr;, 3% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e10% (7% \u0026darr;, 3% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eLow beta (13\u0026ndash;20 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e18% (17% \u0026darr;, 1% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e20% (19% \u0026darr;, 1% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e31% (28% \u0026darr;, 3% \u0026uarr;)*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eHigh beta (21\u0026ndash;30 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e13% (3% \u0026darr;, 10% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e15% (4% \u0026darr;, 11% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e9% (2% \u0026darr;, 7% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eLow gamma (31\u0026ndash;59 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e17% (9% \u0026darr;, 8% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e13% (3% \u0026darr;, 10% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e10% (1% \u0026darr;, 9% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSEG (60\u0026ndash;65 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e11% (1% \u0026darr;, 10% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e17% (1% \u0026darr;, 16% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e20% (1% \u0026darr;, 19% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eHigh gamma (66\u0026ndash;100 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e13% (7% \u0026darr;, 6% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e18% (9% \u0026darr;, 9% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e12% (5% \u0026darr;, 7% \u0026uarr;)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eNot significant\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e0%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e0%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e0%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eTotal \u0026darr; vs. \u0026uarr; physiomarkers\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e48% / 52%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e42% / 58%\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e44% / 56%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\"\u003e\n \u003ch2\u003eDistribution of physiomarkers during the movement state\u003c/h2\u003e\n \u003cp\u003eFor the duration-matched movement dataset, again two peaks were observed in the distribution of physiomarkers (Fig.\u0026nbsp;2B; Table\u0026nbsp;1) with 8% of hemispheres showing the largest effect size in the alpha band and 20% of hemispheres in the low beta band with a MED OFF\u0026thinsp;\u0026gt;\u0026thinsp;MED ON effect peak at 14 Hz. For 17% of hemispheres, the largest effect size was in the SEG band with a MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF effect peak at 62 Hz. For 37% of hemispheres, the largest effect size was found in the theta (9%), high beta (15%), low gamma (13%) or high gamma (18%) band. There was an increased consistency of physiomarker effect directions per frequency band compared to the resting state with clear effect directions in the theta, low-beta and SEG band. A steep increase in the number of SEG physiomarkers was observed (Fig.\u0026nbsp;2B). Median stimulation amplitude was 1.9 mA [1.6, 2.4 mA] for hemispheres with physiomarkers in the alpha-beta band and 2.1 mA [1.7, 2.5 mA] for hemispheres with physiomarkers in the SEG band without a significant difference (MWU test, p\u0026thinsp;\u0026gt;\u0026thinsp;0.10).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\"\u003e\n \u003ch2\u003eDistribution of physiomarkers during the full MDS-UPDRS-III scoring\u003c/h2\u003e\n \u003cp\u003eTo investigate the influence of data recording duration, we recomputed the distribution of physiomarkers based on LFP data acquired during the full MDS-UPDRS-III scoring (3\u0026ndash;7 minutes), which showed two peaks in the alpha/low-beta and SEG band (Fig.\u0026nbsp;2C; Table\u0026nbsp;1). The largest effect of medication was seen in the alpha band for 10% of hemispheres and in the low beta band for 31% of hemispheres, with a MED OFF\u0026thinsp;\u0026gt;\u0026thinsp;MED ON effect peak at 13 Hz. In 35% of hemispheres a MED OFF\u0026thinsp;\u0026gt;\u0026thinsp;MED ON effect direction was seen in the alpha or low beta band. For 20% of hemispheres, the largest effect size was found in the SEG band with a MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF effect peak at 62 Hz. For 27% of hemispheres, the largest effect size was not found in the alpha, low beta or SEG band, but in the theta (8%), high beta (9%), low gamma (10%) or high gamma (12%) band. Compared to the resting state and duration-match movement data, results for the full MDS-UPDRS-III scoring showed the most consistent effect directions across frequency bands with clear patterns emerging in the theta- (MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF), alpha/low-beta- (MED OFF\u0026thinsp;\u0026gt;\u0026thinsp;MED ON), high-beta/low-gamma- (MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF) and SEG- (MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF) band (Fig.\u0026nbsp;2C, Supplementary Figure S2-S4). Using Kruskal-Wallis tests and post-hoc pairwise comparisons with MWU testing on all physiomarker occurrences per frequency band, low beta showed higher occurrence during full MDS-UPDRS-III scoring compared to the resting state (p\u0026thinsp;=\u0026thinsp;0.03), while SEG physiomarkers showed a trend towards more occurrence between these two states (p\u0026thinsp;=\u0026thinsp;0.08). Median stimulation amplitude was 1.9 mA [1.6, 2.4 mA] for hemispheres with physiomarkers in the alpha-beta band and 2.2 mA [1.7, 2.5 mA] for hemispheres with physiomarkers in the SEG band without a significant difference (MWU test).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\"\u003e\n \u003ch2\u003eConsistency of physiomarkers across hemispheres\u003c/h2\u003e\n \u003cp\u003eThe consistency of the effect size direction (e.g., predominantly MED OFF\u0026thinsp;\u0026gt;\u0026thinsp;MED ON or MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF) was statistically tested for the physiomarkers in each frequency band for the resting state, movement state and during the full MDS-UPDRS-III scoring. A high value for the mean sign of the effect sizes (e.g., 1 when all hemispheres show a MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF direction, -1 when all hemispheres show a MED OFF\u0026thinsp;\u0026gt;\u0026thinsp;MED ON direction; versus 0 when effects for the hemispheres are evenly split) was interpreted as a strong directional effect. Per state, binomial tests showed whether a frequency band had a statistically significant predominant effect direction (Table\u0026nbsp;2). Theta (MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF), low beta (MED OFF\u0026thinsp;\u0026gt;\u0026thinsp;MED ON) and SEG (MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF) showed a significant predominant effect direction in all states (p values are reported in Table\u0026nbsp;2). The alpha, high beta, and high gamma band showed no consistent effect direction in either state. For the low gamma band, a significant predominant effect direction (p\u0026thinsp;=\u0026thinsp;0.02, MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF) was found for the full MDS-UPDRS-III testing but not for the other states. When combining the high beta and low gamma band, a common effect direction was found (p\u0026thinsp;=\u0026thinsp;0.00, MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF)\u003c/p\u003e\n \u003cdiv\u003e\n \u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 2\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026ndash; Predominance of physiomarker effect directions across hemispheres per frequency band.\u003c/strong\u003e Mean sign values are reported based on the indicated number of hemispheres (n) for which a physiomarker was detected in the indicated frequency band or combination of frequency bands. Corresponding p-values of a binomial test indicate whether effect directions were significantly more often in the direction of an increase (\u0026uarr;) or a decrease (\u0026darr;) in LFP power with levodopa or were not distinguishable from a 50/50 distribution (-). Findings are listed separately for the resting state, movement state, and full MDS-UPDRS-III scoring.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eFrequency band\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eMean sign (rest)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003eBinomial test\u003c/p\u003e\n \u003cp\u003e(p value)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003eMean sign (movement)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003eBinomial test\u003c/p\u003e\n \u003cp\u003e(p value)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003eMean sign (full MDS-UPDRS-III)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003eBinomial test\u003c/p\u003e\n \u003cp\u003e(p value)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eTheta (4\u0026ndash;8 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e0.87 (n\u0026thinsp;=\u0026thinsp;15)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e0.00*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e0.78 (n\u0026thinsp;=\u0026thinsp;9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e0.04*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e1 (n\u0026thinsp;=\u0026thinsp;8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\n \u003cp\u003e0.00*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eAlpha (9\u0026ndash;12 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e-0.54 (n\u0026thinsp;=\u0026thinsp;13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e0.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e-0.25 (n\u0026thinsp;=\u0026thinsp;8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e0.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e0.4 (n\u0026thinsp;=\u0026thinsp;10)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\n \u003cp\u003e0.34\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eLow beta (13\u0026ndash;20 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e-0.89 (n\u0026thinsp;=\u0026thinsp;18)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e0.00*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e-0.9 (n\u0026thinsp;=\u0026thinsp;20)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e0.00*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e-0.81 (n\u0026thinsp;=\u0026thinsp;31)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\n \u003cp\u003e0.00*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eHigh beta (21\u0026ndash;30 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e0.53 (n\u0026thinsp;=\u0026thinsp;13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e0.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e0.47 (n\u0026thinsp;=\u0026thinsp;15)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e0.11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e0.56 (n\u0026thinsp;=\u0026thinsp;9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\n \u003cp\u003e0.18\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eLow gamma (31\u0026ndash;59 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e-0.06 (n\u0026thinsp;=\u0026thinsp;15)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e0.54 (n\u0026thinsp;=\u0026thinsp;13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e0.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e0.8 (n\u0026thinsp;=\u0026thinsp;10)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\n \u003cp\u003e0.02*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSEG (60\u0026ndash;65 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e0.82 (n\u0026thinsp;=\u0026thinsp;11)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e0.01*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e0.88 (n\u0026thinsp;=\u0026thinsp;17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e0.00*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e0.9 (n\u0026thinsp;=\u0026thinsp;20)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\n \u003cp\u003e0.00*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eHigh gamma (66\u0026ndash;100 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e-0.08 (n\u0026thinsp;=\u0026thinsp;13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e0.00 (n\u0026thinsp;=\u0026thinsp;18)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e0.17 (n\u0026thinsp;=\u0026thinsp;12)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\n \u003cp\u003e0.77\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eAlpha/low-beta (9\u0026ndash;20 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e-0.74 (n\u0026thinsp;=\u0026thinsp;31)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e0.00*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e-0.71 (n\u0026thinsp;=\u0026thinsp;28)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e0.00*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e-0.71 (n\u0026thinsp;=\u0026thinsp;41)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\n \u003cp\u003e0.00*\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eHigh beta/low gamma (21\u0026ndash;59 Hz)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e0.20 (n\u0026thinsp;=\u0026thinsp;30)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e0.36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e0.5 (n\u0026thinsp;=\u0026thinsp;28)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e0.01*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e0.68 (n\u0026thinsp;=\u0026thinsp;19)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\n \u003cp\u003e0.00*\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\u003eWhen combining the adjacent alpha and low beta band, a significant predominant effect direction was found. This can be explained by a general lower frequency of the beta peaks during active DBS as found in previously reported literature comparing the DBS OFF with the DBS ON state (8). Therefore, four distinct effects of medication on LFP power were observed: MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF in theta and high-beta/low-gamma, MED OFF\u0026thinsp;\u0026gt;\u0026thinsp;MED ON in alpha/low-beta and MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF in SEG. Alpha/low-beta and SEG were the frequency bands with most detected physiomarkers.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec17\"\u003e\n \u003ch2\u003eClinical characteristics associated with physiomarker effect size and frequency\u003c/h2\u003e\n \u003cp\u003eTo investigate the association between clinical features and the presence of either an alpha/low-beta or SEG physiomarker across hemispheres and their magnitude of fluctuation, a k-means clustering analysis was performed using the mean of all 1-Hz bin effect sizes in the alpha/low beta and SEG band separately, stimulation amplitude (in mA), levodopa dose (in mg) and MDS-UPDRS-III response (in points reduction) to levodopa (five variables in total). Two hemispheres could not be included in the analysis due to incomplete MDS-UPDRS-III scores. The median silhouette score indicated a moderate structure for the optimum of four clusters (0.38, Supplementary Fig.\u0026nbsp;6) (27). The separation between the four clusters is visualized in Fig.\u0026nbsp;3 using three principal components that accounted for over 70% of the variance, a common method for visualizing clusters in a reduced-dimension space when multiple variables are used (28). For each cluster, we counted the number of hemispheres with physiomarkers in the alpha/low-beta and SEG band. Alpha/low-beta physiomarkers predominated cluster 1 and 3 (eleven out of twelve and eight out of nine hemispheres respectively), while SEG physiomarkers predominated cluster 2 (seven out of seven hemispheres). Cluster 4 contained a mixture of 14 hemispheres with alpha/low-beta physiomarkers and 10 hemispheres with SEG physiomarkers out of 24 hemispheres.\u003c/p\u003e\n \u003cp\u003eStatistical evaluation using one-way ANOVA indicated that several features differed significantly across clusters (all p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, see Table 3). Pairwise testing with post-hoc Tukey\u0026rsquo;s HSD tests (Table 4) revealed a lower alpha/low-beta average effect size per hemisphere in cluster 1 compared to all other clusters and a lower alpha/low-beta average effect size per hemisphere in cluster 3 compared to cluster 2. Higher SEG average effect size per hemisphere was observed in cluster 2 compared to all other clusters. Cluster 3 had higher levodopa doses than cluster 2 and 4, cluster 1 contained lower stimulation amplitudes than cluster 2 and 3 and cluster 2 contained lower reductions in MDS-UPDRS-III scores following levodopa intake than cluster 1 and 4 but not compared to cluster 3. Broadly speaking, we can hence distinguish the following neurophysiological and clinical profiles: hemispheres in cluster 1 contained primarily alpha/low-beta physiomarkers, low stimulation amplitudes, average levodopa doses and a strong MDS-UPDRS-III response to levodopa; all hemispheres in cluster 2 contained SEG physiomarkers and showed relatively high stimulation amplitude, average levodopa dose and a weak MDS-UPDRS-III response to levodopa; hemispheres in cluster 3 contained primarily alpha/low-beta physiomarkers, high stimulation amplitudes, high doses and weak MDS-UPDRS-III responses to levodopa; cluster 4 (the largest group) contained a mixture of hemispheres with either alpha/low-beta and SEG physiomarkers and also showed the least distinctive clinical features. Stimulation amplitudes were relatively low compared cluster 3 (p\u0026thinsp;=\u0026thinsp;0.02) and cluster 2 (p\u0026thinsp;=\u0026thinsp;0.05). Levodopa doses were lower than cluster 2 and 3 (p\u0026thinsp;=\u0026thinsp;0.00) but not compared to cluster 1, and MDS-UPDRS-III response to levodopa were weaker than cluster 1 (p\u0026thinsp;=\u0026thinsp;0.00) but stronger than cluster 2 and 3 (p\u0026thinsp;=\u0026thinsp;0.01, p\u0026thinsp;=\u0026thinsp;0.01, respectively).\u003c/p\u003e\n \u003cdiv\u003e\n \u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 3\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026ndash; Clinical characteristics of identified clusters and ANOVA results\u003c/strong\u003e Average alpha/low-beta and SEG effect size per hemisphere, Stimulation amplitude (in mA), Dose (in mg), MDS-UPDRS-III response (in points reduction) were determined per cluster after k-means clustering. ANOVA analyses revealed significant differences in these metrics across clusters.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\n \u003cp\u003eAlpha/low-beta effect size per hemisphere (median, IQR)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eCluster 1\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003eCluster 2\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003eCluster 3\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003eCluster 4\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003eANOVA\u003c/p\u003e\n \u003cp\u003e(p-value)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003eAll clusters\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e-0.57 [-0.71, -0.44]\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e-0.01 [-0.22, 0.02]\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e-0.36 [-0.51, -0.17]\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e-0.17 [-0.28, -0.08]\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.00\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e-0.26 [-0.44, -0.10]\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSEG effect size per hemisphere (median, IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e0.25 [0.13, 0.46]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e0.97 [0.72, 1.29]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e0.16 [-0.02, 0.29]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e0.3 [0.14, 0.45]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e0.31 [0.14, 0.57]\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eStimulation (mA, median, IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e1.7 [1.4, 2.0]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e2.5 [2.0, 2.6]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e2.4 [2.1, 2.8]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e1.9 [1.5, 2.3]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e2.0 [1.6, 2.4]\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eDose (mg, median, IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e150 [125, 200]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e150 [150, 150]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e200 [175, 350]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e150 [100, 150]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e150 [150, 188]\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eMDS-UPDRS-III response (points, median, IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e-14 [-18, -11]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e-5 [-6, -4]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e-4 [-9, -3]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e-10 [-12, -8]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e-10 [-12, -6]\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cdiv\u003e\n \u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 4\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003e\u0026ndash; Summary of post-hoc pair-wise statistical testing, using the Tukey HSD test.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eAlpha/low-beta effect size,\u003c/p\u003e\n \u003cp\u003eSEG effect size,\u003c/p\u003e\n \u003cp\u003eStimulation,\u003c/p\u003e\n \u003cp\u003eDose,\u003c/p\u003e\n \u003cp\u003eLevodopa MDS-UPDRS-III response\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003eCluster 1\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003eCluster 2\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003eCluster 3\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003eCluster 4\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eCluster 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003ex\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.00,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.00,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.01,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.00,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.00,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.76,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.00,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.86,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.00,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;1,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.70,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.11,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eCluster 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003ex\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003eX\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.00,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.00,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.99,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.00,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.31,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.00,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.05,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.00,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eCluster 3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003ex\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003eX\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003ex\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.09,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.68,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.02,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.00,\u003c/p\u003e\n \u003cp\u003ep\u0026thinsp;=\u0026thinsp;0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eCluster 4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003ex\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003eX\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003ex\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003ex\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003ctfoot\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\"\u003eAverage alpha/low-beta and SEG effect size per hemisphere, stimulation amplitude (in mA), levodopa dose (in mg), MDS-UPDRS-III response (in points reduction) were pair-wise compared between clusters post-hoc using Tukey\u0026rsquo;s Honestly Significant Difference after ANOVA analyses revealed significant differences between clusters. The Tukey\u0026rsquo;s HSD allowed for multiple comparisons while controlling the probability of making one or more Type I errors.\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tfoot\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cdiv\u003e\u003c/div\u003e\n \u003cp\u003eTo summarize, alpha/low-beta physiomarkers in cluster 1 were associated with a relatively strong clinical response to levodopa despite active DBS. SEG physiomarkers in cluster 2 were associated with relatively small dose of medication and a small clinical response but a relatively high stimulation amplitude. Alpha/low-beta physiomarkers in cluster 3 were associated with a high levodopa dose but without a strong clinical effect (12). In cluster 4, hemispheres either contained an alpha/low-beta physiomarker or an SEG physiomarker without a clear predominance of the two. Therefore, this cluster did not reveal a distinct clinical profile relating to either alpha/low-beta or SEG physiomarkers. It is still unclear if some of these hemispheres could show an SEG physiomarker if stimulation amplitude increased, as entrainment has been shown to be a nonlinear process with patient-specific thresholds (29).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec18\"\u003e\n \u003ch2\u003eFull spectrum response to levodopa (2-100 Hz) differs according to physiomarker frequency\u003c/h2\u003e\n \u003cp\u003eTo further investigate the specificity of levodopa-induced spectral changes for the different physiomarker frequencies, the power values from the full spectrum (2-100 Hz) for all segments per hemisphere grouped by the physiomarker frequency band (theta, alpha/low-beta, high beta, low gamma, SEG or high gamma physiomarkers) were compared between OFF- and ON-MED with permutation testing (Supplementary Fig.\u0026nbsp;5). This figure shows that hemispheres with a physiomarker in the theta, high-beta, low-gamma and high gamma band, displayed broad spectral power changes after levodopa intake without clear modulation of low-beta or SEG. By contrast, no clear OFF-related peak in the beta band was observed for hemispheres with an SEG physiomarker. In the hemispheres with a physiomarker in either the alpha/low-beta or SEG band, a clear peak was seen centering around 62,5 Hz in the MED ON state, which reduced in the MED OFF state (29). In summary, depending on physiomarker frequency, distinct levodopa-induced changes were observed across the power spectrum suggesting different neurophysiological responses.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we investigated the consistency of levodopa-induced changes in STN LFP power. A data-driven method was used to identify a 1-Hz frequency band in the STN LFP spectrum per hemisphere with the largest effect size when transitioning from the OFF to ON levodopa state during active DBS. We determined these \u0026lsquo;physiomarkers\u0026rsquo; in three different states (30\u0026ndash;60 seconds rest, randomly picked 30\u0026ndash;60 seconds movement, and the full (3\u0026ndash;7 minutes) MDS-UPDRS-III scoring) to investigate the effect of movement on physiomarker frequency, as beta oscillations have been shown to shift frequency with active movement (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e), and identify if longer periods of data resulted in more stable levodopa-related spectral changes (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Supplementary Figure S2-S4). The MDS-UPDRS-III score represents common daily movements and was used to assess the effect of levodopa and DBS in our center and, thus, representative of remaining treatment fluctuations that may be treated by aDBS. The most consistent set of physiomarkers when observing all frequency bands (i.e. consistent effect directions in the theta, alpha, low beta, high beta, low gamma, SEG and high gamma band) was extracted from LFPs recorded during the full MDS-UPDRS-III scoring (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAlthough across-hemisphere analyses did not reveal significant power changes across the broad frequency spectrum when comparing the OFF- to the ON levodopa state (Supplementary Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e), the within-hemisphere analyses revealed a personalised, statistically significant, levodopa-related physiomarker in every hemisphere. Detected physiomarkers most commonly had a frequency in the alpha/low-beta band (9\u0026ndash;20 Hz, 41% of hemispheres), primarily showing suppression of power in the ON-dopaminergic compared to OFF-dopaminergic state (35% of hemispheres). The second most common physiomarker frequency band was SEG (20% of hemispheres), primarily showing an increase in power towards the ON-dopaminergic state compared with the OFF-dopaminergic state (19% of hemispheres). A slightly smaller percentage of hemispheres (19%) showed consistent increases in broadband high beta to low gamma power, while a minority of hemispheres had a physiomarker in the theta band (8%). In 12% of hemispheres the physiomarker frequency was found in the high gamma band but with conflicting effect directions.\u003c/p\u003e \u003cp\u003eNext, we performed a k-means clustering analysis to identify clinical characteristics associated with the effect size of levodopa-induced spectral changes in the alpha/low-beta and SEG band. Four clusters were detected, for which two clusters linked the occurrence of alpha/low-beta physiomarkers with a relatively strong clinical response to levodopa or a high levodopa dose but weak clinical response, respectively. A third cluster linked the occurrence of SEG physiomarkers with a relatively small levodopa dose and clinical response but high stimulation amplitudes. However, the majority of hemispheres were assigned to a fourth cluster that contained physiomarkers in both the alpha/low-beta and SEG band, precluding any associations for these hemispheres with clinical variables. An important consideration here is that entrainment leading to an SEG physiomarker may depend on patient-specific stimulation thresholds (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e). Therefore, some of these hemispheres may not have reached the appropriate stimulation amplitude for entrainment (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e). Finally, while there was some SEG activity in hemispheres that showed an alpha/low-beta physiomarker, alpha/low-beta activity was negligible in hemispheres that showed an SEG physiomarker (Supplementary Figure S5). This corresponds with literature (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e), showing minimal contribution of beta power to distinguishing the OFF- from the ON-dopaminergic state when SEG is present. Alpha/low-beta physiomarkers can be used in the currently available aDBS system (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e), while SEG physiomarkers cannot. Therefore, our results support the use of alpha/low-beta based aDBS based upon levodopa fluctuations in a sizable number of patients (35% of our sample). This is in contrast with several, smaller, reports that found beta to be a poor physiomarker for aDBS (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan additionalcitationids=\"CR32\" citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn the resting state, there was less consistency across hemispheres of physiomarker effect directions in the high-beta and low-gamma band compared to the movement state, possibly related to aperiodic activity (Supplementary Figure S5) (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan additionalcitationids=\"CR35 CR36\" citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e). Typically, in this frequency band no distinct spectral peaks are observed but the power follows a characteristic 1/f-like pattern that is associated with aperiodic neural activity. The direction of the medication effect showed an increase in power after levodopa intake, which was also found in the paper by Gerster et al. (\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e). Furthermore, we found during the full MDS-UPDRS-III scoring that theta frequencies increased with power after levodopa intake (8% of hemispheres), similar to high-beta and low-gamma oscillations. The only commercially available aDBS system (\u0026lsquo;Percept\u003csup\u003eTM\u003c/sup\u003e\u0026rsquo;) currently does not correct for aperiodic activity when adapting the stimulation based on the power of a physiomarker. However, using long periods (minutes) for measuring physiomarker activity might result in more reliable estimates of the medication state (corresponding with our results regarding the consistency of physiomarker distributions in different states, Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). This may explain the significant improvement of ON-time duration without troublesome dyskinesia in the slower, dual-threshold (DT) aDBS compared to continuous DBS (cDBS) that was observed in the ADAPT PD trial while in the faster, single-threshold (ST) aDBS this was not observed (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe majority of physiomarkers were found in the alpha/low-beta or SEG band. The occurrence of similar effect directions in the alpha and low-beta band may point towards similar neural processes, possibly explained by the frequently observed shift in beta frequencies when stimulation is turned on (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e). Therefore, determining the physiomarker during stimulation, but also during movement, is key for picking the right frequency band used for aDBS (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). To generalize these findings to the clinical application of aDBS, longitudinal recordings are necessary to investigate the consistency of the personalized physiomarker to treatment alterations and symptom burden. For low beta and SEG power, this has already been established in previous work (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). However, for theta or broadband high beta to low gamma power this has not been investigated yet (Supplementary Fig.\u0026nbsp;5A, C, D). The contribution of theta, alpha and high beta in explaining the variance in patients' hemibody impairment OFF medication and the change in hemibody impairment following levodopa was reported previously, however, without active DBS and without showing the marked inter-individual differences in physiological responses to levodopa in PD that we show in this study (\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e). It is unknown if our results are applicable to GPi recordings, because to our knowledge, no clear modulation of beta power to levodopa has been shown to date. Furthermore, multiple factors may influence levodopa-related LFP physiomarker fluctuations, especially sleep, movement and stress can affect beta power (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e). These effects and the physiomarkers that respond to them should be further investigated to determine the optimal feedback signal for aDBS.\u003c/p\u003e \u003cp\u003eAlthough this study did not examine symptom specificity in relation to changes in specific frequency bands, it is plausible that a personalized model of STN activity\u0026mdash;across different clinical states and accounting for both rest and movement\u0026mdash;could more accurately inform adaptive DBS compared to current approaches that rely solely on band-based frequency features (\u003cspan additionalcitationids=\"CR43 CR44\" citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e). Recent machine learning studies already showed marked improvement in classification of the OFF- and ON-dopaminergic state and motor performance when including the full frequency spectrum and burst time-dynamics (\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e). Future research could consider deep learning techniques that are already being applied for longitudinal Parkinson\u0026rsquo;s disease severity monitoring using other types of data, such as kinematics (\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e).\u003c/p\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eLimitations\u003c/h2\u003e \u003cp\u003eOur data-driven method for determining personalized physiomarkers of the dopaminergic state during active DBS revealed that the neurophysiological response to levodopa varies substantially between individuals. However, there are some caveats arising from the methods we used. To start, the short segment of data (2 seconds) used for the calculation of the power spectral densities (PSDs) may have led to spectral leakage and insufficient frequency resolution for accurately characterizing narrowband oscillatory phenomena, such as SEG, and peak frequency estimation. Shorter time segments, however, provide greater temporal specificity for capturing transient burst dynamics that respond to levodopa, particularly in the beta band. Currently the only commercially approved Percept\u003csup\u003e\u0026trade;\u003c/sup\u003e system for aDBS uses a similar frequency resolution (0.98 Hz) and PSD overlap (80%) for the dual threshold algorithm and sensing durations for altering stimulation amplitude between 1 and 2 seconds (\u0026lsquo;Onset duration\u0026rsquo;), corresponding largely with the PSD calculation in this study (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e). Also, the configuration of the DBS stimulation may have been suboptimal in some patients despite completing cDBS titration, possibly causing the levodopa-related spectral fluctuations to be different and less disease-specific than those from the optimal contact used for stimulation.\u003c/p\u003e \u003cp\u003eThe heterogeneity in the duration of the recordings across patients during full MDS-UPDRS-III testing may have affected the accuracy of some physiomarkers, since this duration varied from about 3 to 7 minutes. Nevertheless, several minutes of data would still provide more than 100 PSDs per hemisphere for the Cohen\u0026rsquo;s d effect size computation. Besides technical factors, the variability in symptom profiles across patients likely influenced the detection of a physiomarker in the alpha/low-beta or SEG bands, which was not corrected or stratified for. For example, patients with tremor-predominant phenotypes tend to show a weaker association between LFP frequency bands and dopaminergic state compared to those with bradykinesia- or rigidity-predominant symptoms (\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study is the largest to date (n\u0026thinsp;=\u0026thinsp;100 hemispheres) to investigate the frequency-specificity of levodopa-induced changes in STN spectral power at an individual level. We found that for 35% of hemispheres, the largest change in 1-Hz bin power after levodopa intake, referred to as \u0026ldquo;physiomarker\u0026rdquo;, was a reduction in power in the alpha/low-beta band, while 19% had an increase in SEG. Remaining physiomarkers were found in other frequency bands. Therefore, aDBS that can be tailored according to the individual neurophysiological profile beyond solely beta frequencies may improve aDBS outcome. The majority (56%) of physiomarkers exhibited a MED ON\u0026thinsp;\u0026gt;\u0026thinsp;MED OFF effect after levodopa intake, which is incompatible with current commercially-approved aDBS software. The neurophysiological signature identifying the dopaminergic state during active stimulation varies substantially between individuals and predicting the right physiomarker with just clinical characteristics may not be possible for all patients. This opens the door to machine-learning\u0026ndash;based identification of disease states for aDBS and calls for commercial software that enables stimulation control based on multiple, and potentially inverse, LFP frequency features.\u003c/p\u003e "},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData is not available for public sharing due to ethical restrictions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe underlying code for this study is not publicly available but may be made available to qualified researchers on reasonable request from the corresponding author.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eMGJdN contributed to the design, execution, data acquisition, analysis, writing and editing of the final manuscript. BK, DH and MJS contributed to the data acquisition and editing of the final manuscript. RMAdB and RS contributed to the design and editing of the final manuscript. CRO, BCMvW, AWB and MB contributed to the design, writing and editing of the final manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the Amsterdam UMC TKI-PPP grant (2021 \u0026amp; 2023 call, Innovation Exchange Amsterdam).\u0026nbsp;The funder played no role in study design, data collection, analysis and interpretation of data, or the writing of this manuscript.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRS is consultant in the educational programs of Medtronic and Boston Scientific\u003c/p\u003e\n\u003cp\u003eMB additionally received research funding related to the manuscript from Medtronic (2023, 2024).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eMGJdN, MJS, BK, DH, AWB, BCMvW, RMAdB have no conflicts of interests to declare in relation to the manuscript.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eLimousin, P., Foltynie, T. Long-term outcomes of deep brain stimulation in Parkinson disease. Nature Reviews Neurology. 2019;15(4):234\u0026ndash;42.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMansouri, A. et al. Deep brain stimulation for Parkinson\u0026rsquo;s disease: meta-analysis of results of randomized trials at varying lengths of follow-up. Journal of Neurosurgery JNS. 2018;128(4):1199\u0026ndash;213.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDeuschl, G. et al. European Academy of Neurology/Movement Disorder Society - European Section guideline on the treatment of Parkinson's disease: I. Invasive therapies. Eur J Neurol. 2022;29(9):2580\u0026ndash;95.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHauser, R.A., McDermott, M.P., Messing, S. Factors associated with the development of motor fluctuations and dyskinesias in Parkinson disease. Arch Neurol. 2006;63(12):1756\u0026ndash;60.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWeaver, F.M. et al. Bilateral deep brain stimulation vs best medical therapy for patients with advanced Parkinson disease: a randomized controlled trial. Jama. 2009;301(1):63\u0026ndash;73.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFollett, K.A. et al. Pallidal versus subthalamic deep-brain stimulation for Parkinson's disease. N Engl J Med. 2010;362(22):2077\u0026ndash;91.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOdekerken, V.J. et al. Subthalamic nucleus versus globus pallidus bilateral deep brain stimulation for advanced Parkinson's disease (NSTAPS study): a randomised controlled trial. Lancet Neurol. 2013;12(1):37\u0026ndash;44.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMathiopoulou, V. et al. Modulation of subthalamic beta oscillations by movement, dopamine, and deep brain stimulation in Parkinson\u0026rsquo;s disease. npj Parkinson's Disease. 2024;10(1):77.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evan Wijk, B.C.M., de Bie, R.M.A., Beudel, M. A systematic review of local field potential physiomarkers in Parkinson's disease: from clinical correlations to adaptive deep brain stimulation algorithms. J Neurol. 2023;270(2):1162\u0026ndash;77.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen, Y. et al. Neuromodulation effects of deep brain stimulation on beta rhythm: A longitudinal local field potential study. Brain Stimul. 2020;13(6):1784\u0026ndash;92.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLofredi, R. et al. Subthalamic beta bursts correlate with dopamine-dependent motor symptoms in 106 Parkinson\u0026rsquo;s patients. npj Parkinson's Disease. 2023;9(1):2.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ede Neeling, M.G.J. et al. Beta Power Response After Levodopa Intake in Parkinson's Disease Patients With Chronic Sensing-Enabled Deep Brain Stimulation. Annals of Neurology. 2025;n/a(n/a).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBelova, E.M., Semenova, U., Gamaleya, A.A., Tomskiy, A.A., Sedov, A. Voluntary movements cause beta oscillations increase and broadband slope decrease in the subthalamic nucleus of parkinsonian patients. Eur J Neurosci. 2021;53(7):2205\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOehrn, C.R. et al. Chronic adaptive deep brain stimulation versus conventional stimulation in Parkinson\u0026rsquo;s disease: a blinded randomized feasibility trial. Nature Medicine. 2024;30(11):3345\u0026ndash;56.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSermon, J.J. et al. Sub-harmonic entrainment of cortical gamma oscillations to deep brain stimulation in Parkinson's disease: Model based predictions and validation in three human subjects. Brain Stimul. 2023;16(5):1412\u0026ndash;24.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWiest, C., et al. Finely-tuned gamma oscillations: Spectral characteristics and links to dyskinesia. Exp Neurol. 2022;351:113999.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evon Elm, E., Altman, D.G., Egger, M., Pocock, S.J., G\u0026oslash;tzsche, P.C., Vandenbroucke, J.P. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007;370(9596):1453\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStam, M.J. et al. AI-DBS study: protocol for a longitudinal prospective observational cohort study of patients with Parkinson's disease for the development of neuronal fingerprints using artificial intelligence. BMJ Open. 2025;15(5):e091563.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHolewijn, R.A. et al. General Anesthesia vs Local Anesthesia in Microelectrode Recording-Guided Deep-Brain Stimulation for Parkinson Disease: The GALAXY Randomized Clinical Trial. JAMA Neurol. 2021;78(10):1212\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGoyal, A. et al. The development of an implantable deep brain stimulation device with simultaneous chronic electrophysiological recording and stimulation in humans. Biosens Bioelectron. 2021;176:112888.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOostenveld, R., Fries, P., Maris, E., Schoffelen, J.M. FieldTrip: Open Source Software for Advanced Analysis of MEG, EEG, and Invasive Electrophysiological Data. Computational Intelligence and Neuroscience. 2011;2011(1):156869.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStam, M.J. et al. A comparison of methods to suppress electrocardiographic artifacts in local field potential recordings. Clinical Neurophysiology. 2023;146:147\u0026ndash;61.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMaris, E., Oostenveld, R. Nonparametric statistical testing of EEG- and MEG-data. Journal of Neuroscience Methods. 2007;164(1):177\u0026ndash;90.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGerber, E.M. permutest Retrieved November 11, 2024 ed: (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.mathworks.com/matlabcentral/fileexchange/71737-permutest\u003c/span\u003e\u003cspan address=\"https://www.mathworks.com/matlabcentral/fileexchange/71737-permutest\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), MATLAB Central File Exchange.; 2024.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLakens, D. Calculating and reporting effect sizes to facilitate cumulative science: a practical primer for t-tests and ANOVAs. Front Psychol. 2013;4:863.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchinkel, M., Appelman, B., Butler, J., Schuurman, A., Wiersinga, W.J. Association of clinical sub-phenotypes and clinical deterioration in COVID-19: further cluster analyses. Intensive Care Med. 2021;47(4):482\u0026ndash;4.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRousseeuw, P.J. Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. Journal of Computational and Applied Mathematics. 1987;20:53\u0026ndash;65.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDing, C., He, X., editors. K-means clustering via principal component analysis. Proceedings of the twenty-first international conference on Machine learning; 2004.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMathiopoulou, V. et al. Gamma entrainment induced by deep brain stimulation as a biomarker for motor improvement with neuromodulation. Nature Communications. 2025;16(1):2956.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStanslaski, S. et al. Sensing data and methodology from the Adaptive DBS Algorithm for Personalized Therapy in Parkinson\u0026rsquo;s Disease (ADAPT-PD) clinical trial. npj Parkinson's Disease. 2024;10(1):174.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eColombo, A. et al. Finely Tuned γ Tracks Medication Cycles in Parkinson's Disease: An Ambulatory Brain-Sense Study. Mov Disord. 2025;40(5):881\u0026ndash;95.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGerster, M. et al. Beyond beta rhythms: Aperiodic broadband power reflects Parkinson\u0026rsquo;s disease severity\u0026ndash;a multicenter study. bioRxiv. 2025:2025.03.11.642600.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBronte-Stewart H.M. et al. Long-Term Personalized Adaptive Deep Brain Stimulation in Parkinson Disease: A Nonrandomized Clinical Trial. JAMA Neurology. 2025.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWiest, C. et al. The aperiodic exponent of subthalamic field potentials reflects excitation/inhibition balance in Parkinsonism. Elife. 2023;12.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStam, M.J. et al. Long-term consistency of aperiodic and periodic physiomarkers in subthalamic local field potentials in Parkinson\u0026rsquo;s disease. npj Parkinson's Disease. 2025;11(1):204.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim, J. et al. Dopamine depletion can be predicted by the aperiodic component of subthalamic local field potentials. Neurobiology of Disease. 2022;168:105692.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGerster, M. et al. Beyond beta rhythms: subthalamic aperiodic broadband power scales with Parkinson's disease severity-a cross-sectional multicentre study. EBioMedicine. 2025;122:105988.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeldmann, L.K. et al. Toward therapeutic electrophysiology: beta-band suppression as a biomarker in chronic local field potential recordings. npj Parkinson's Disease. 2022;8(1):44.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKhawaldeh, S. et al. Balance between competing spectral states in subthalamic nucleus is linked to motor impairment in Parkinson's disease. Brain. 2022;145(1):237\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaumgartner, A.J., Hirt, L., Amara, A.W., Kern, D.S., Thompson, J.A. Diurnal fluctuations of subthalamic nucleus local field potentials follow naturalistic sleep-wake behavior in Parkinson\u0026rsquo;s disease. Sleep. 2025;48(6):zsaf005.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeldmann LK, Lofredi R, Al-Fatly B, Busch JL, Mathiopoulou V, Roediger J, et al. Christmas-Related Reduction in Beta Activity in Parkinson's Disease. Mov Disord. 2023;38(4):692\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMerk, T. et al. Machine learning based brain signal decoding for intelligent adaptive deep brain stimulation. Experimental Neurology. 2022;351:113993.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSand, D. et al. Machine learning-based personalized subthalamic biomarkers predict ON-OFF levodopa states in Parkinson patients. J Neural Eng. 2021;18(4).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePeterson, V. et al. Movement decoding using spatio-spectral features of cortical and subcortical local field potentials. Exp Neurol. 2023;359:114261.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eToth, R. et al. DyNeuMo Mk-2: An Investigational Circadian-Locked Neuromodulator with Responsive Stimulation for Applied Chronobiology. Conf Proc IEEE Int Conf Syst Man Cybern. 2020;2020:3433\u0026ndash;40.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCasta\u0026ntilde;o-Candamil, S. et al. Identifying controllable cortical neural markers with machine learning for adaptive deep brain stimulation in Parkinson's disease. Neuroimage Clin. 2020;28:102376.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFranco, A. et al. The Role of Deep Learning and Gait Analysis in Parkinson\u0026rsquo;s Disease: A Systematic Review. Sensors. 2024;24(18):5957.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeudel, M. et al. Oscillatory Beta Power Correlates With Akinesia-Rigidity in the Parkinsonian Subthalamic Nucleus. Mov Disord. 2017;32(1):174\u0026ndash;5.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"npj-parkinsons-disease","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjparkd","sideBox":"Learn more about [npj Parkinson's Disease](http://www.nature.com/npjparkd/)","snPcode":"41531","submissionUrl":"https://submission.springernature.com/new-submission/41531/3","title":"npj Parkinson's Disease","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-9190790/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9190790/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAdaptive deep brain stimulation (aDBS) can adjust stimulation to the (medication) state of patients with Parkinson\u0026rsquo;s disease using physiological biomarkers (physiomarkers), but whether a universal physiomarker frequency exists remains unknown. We quantified levodopa-induced spectral changes (2\u0026ndash;100 Hz) during active DBS before and after levodopa intake in local field potential recordings from 52 PD patients (100 hemispheres). Using cluster-based permutation statistics, we assigned the frequency with the largest significant change as the personalized physiomarker. Alpha/low-beta (9\u0026ndash;20 Hz) power reduction was the most common physiomarker (35% of hemispheres) followed by stimulation-entrained gamma (SEG, 62.5 Hz) power increases (19% of hemispheres). Strong levodopa responses, high levodopa doses, and lower stimulation amplitudes were mostly associated with alpha/low-beta physiomarkers, while higher stimulation amplitudes and average levodopa doses were associated with SEG physiomarkers. These findings argue against a single generalizable physiomarker frequency and support individualized selection based on patient-specific levodopa and stimulation profiles.\u003c/p\u003e","manuscriptTitle":"Levodopa-related physiomarkers during deep brain stimulation in Parkinson’s disease","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-07 09:09:26","doi":"10.21203/rs.3.rs-9190790/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-04-20T12:05:45+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-20T07:03:29+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-12T03:55:37+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"305241260638405320500690144425579639793","date":"2026-04-07T23:52:20+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-07T13:03:21+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"251834135829928686605798432485777359776","date":"2026-04-03T09:30:58+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-02T09:09:41+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"17695278751536461475867460008368935742","date":"2026-04-01T06:17:52+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"293611779995942239128381618627961422655","date":"2026-03-31T14:47:23+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-03-31T14:33:28+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-26T12:49:58+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-26T09:24:00+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Parkinson's Disease","date":"2026-03-22T11:15:37+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-parkinsons-disease","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjparkd","sideBox":"Learn more about [npj Parkinson's Disease](http://www.nature.com/npjparkd/)","snPcode":"41531","submissionUrl":"https://submission.springernature.com/new-submission/41531/3","title":"npj Parkinson's Disease","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"29b2c6c0-10e1-4fbb-820c-3d368bde29a8","owner":[],"postedDate":"April 7th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":65670399,"name":"Health sciences/Biomarkers"},{"id":65670400,"name":"Health sciences/Medical research"},{"id":65670401,"name":"Health sciences/Neurology"},{"id":65670402,"name":"Biological sciences/Neuroscience"}],"tags":[],"updatedAt":"2026-05-17T09:23:57+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-07 09:09:26","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9190790","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9190790","identity":"rs-9190790","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}