Predicting Neuromodulation Outcome for Parkinson's Disease with Generative Virtual Brain Model

Siyuan Du; Siyi Li; Shuwei Bai; Ang Li; Haolin Li; Mingqing Xiao; Yang Pan; Dongsheng Li

q-bio.NCMar 31, 2026

Predicting Neuromodulation Outcome for Parkinson's Disease with Generative Virtual Brain Model

Authors: Siyuan Du, Siyi Li, Shuwei Bai, Ang Li, Haolin Li, Mingqing Xiao, Yang Pan, Dongsheng Li,

Organizations: College of Computer Science and Artificial Intelligence, Fudan University, Shanghai, China · Shanghai AI Laboratory, Shanghai, China · University of Science and Technology of China, Anhui, China · Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China ·

Abstract

Parkinson's disease (PD) affects over ten million people worldwide. Although temporal interference (TI) and deep brain stimulation (DBS) are promising therapies, inter-individual variability limits empirical treatment selection, increasing non-negligible surgical risk and cost. Previous explorations either resort to limited statistical biomarkers that are insufficient to characterize variability, or employ AI-driven methods which is prone to overfitting and opacity. We bridge this gap with a pretraining-finetuning framework to predict outcomes directly from resting-state fMRI. Critically, a generative virtual brain foundation model, pretrained on a collective dataset (2707 subjects, 5621 sessions) to capture universal disorder patterns, was finetuned on PD cohorts receiving TI (n=51) or DBS (n=55) to yield individualized virtual brains with high fidelity to empirical functional connectivity (r=0.935). By constructing counterfactual estimations between pathological and healthy neural states within these personalized models, we predicted clinical responses (TI: AUPR=0.853; DBS: AUPR=0.915), substantially outperforming baselines. External and prospective validations (n=14, n=11) highlight the feasibility of clinical translation. Moreover, our framework provides state-dependent regional patterns linked to response, offering hypothesis-generating mechanistic insights.

Figures & tables

Figure 1: Clinical gap, proposed framework, and cohort statistics. a, Overview of trial-and-error neuromodulation process for PD and the clinical situation . b, Overview of the proposed framework: a generative virtual brain paradigm transfers dynamical priors from large-scale data to small clinical cohorts for individualized neuromodulation response prediction. c–d, Statistics of the pretraining cohort: sample distribution across four centers ( c ) and disease spectrum composition ( d ). e–g, Statistics of the Ruijin PD cohorts: age distribution stratified by sex ( e ), MDS-UPDRS III motor scores distribution ( f ) and PD severity composition ( g ). h, Image processing pipeline: from raw BOLD acquisition to parcellated regional time series.

Figure 2: Foundation and individualized virtual brain validation. a, Architecture of the generative virtual brain model. b, BOLD signal forecasting in four Parkinson’s disease–relevant regions on the test set and global MAE across 166 regions: ANN = 0.7044, FVB = 0.6576, iVB = 0.5346. c, Per-region R 2 distribution across 166 brain areas. d, Cumulative distribution function (CDF) of R 2 values. e–g, Functional connectivity validation for ANN ( e ; $r=0.630$ ), FVB ( f ; $r=0.867$ ), and iVB ( g ; $r=0.948$ ) against empirical data (all $p<10^{-200}$ , two-tailed Pearson correlation). h, Visualization of brain functional connectivity using Yeo 7-network parcellation illustrates that iVB preserves high-level connectivity patterns closely matching empirical data. i, Boxplots w.r.t. the performance comparison between FVB and iVB across 106 patients (TI and DBS cohorts) show the significant improvements in MAE, R 2 , and FC correlation after finetuning (all $p<0.001$ , two-tailed paired t-test).

Figure 3: Validation of predicting neuromodulation response. a, Schematic illustration of the iVB-based workflow for predicting neuromodulation response. b, Diagram depicting the calculation of the Counterfactual Brain Mismatch. c, Clinical variable comparisons between responders/non-responders in TI ( $n=51$ ) and DBS ( $n=55$ ) cohorts. Fisher’s exact test for sex (Male/Female) and target (GPi/STN); Mann–Whitney U test for age. All comparisons non-significant (ns; $p>0.05$ ). d–f, Five-fold cross-validation performance comparing the proposed iVB-based model (Ours) with baselines. d, Summary metrics (mean $\pm$ 95% CI) for Accuracy (ACC), F1, and Precision. Ours achieved strong performance in both TI (ACC = 0.845, F1 = 0.816, Precision = 0.843) and DBS (ACC = 0.836, F1 = 0.847, Precision = 0.917), significantly outperforming all baselines (paired $t$ -tests, all $p<0.05$ ). Curves show mean ROC and PR with 95% CI. e, TI cohort: AUC = $0.823\pm 0.083$ , AUPR = $0.853\pm 0.066$ (ours). f, DBS cohort: AUC = $0.865\pm 0.072$ , AUPR = $0.915\pm 0.048$ (ours). g–h, Forest plots report subgroup-wise AUC (circles) and AUPR (squares) across five cross-validation folds for the TI ( g ) and DBS ( h ). Logistic interaction tests showed no significant effect modification (all $p>0.05$ )

Figure 4: Clinical utility and validation of the iVB-based prediction framework. a, Decision curve analysis (DCA) for the proposed model in TI and DBS cohorts. Our model (solid lines) outperforms reference strategies "Treat all" (dashed blue) and "Treat none" (dashed yellow). b, Raincloud plots of predicted response probabilities stratified by clinical response status (Responders vs. Non-responders). Statistical comparisons: TI cohort (Mann-Whitney $U=33.00$ , $p<0.001$ , Cohen’s $d=2.39$ ); DBS cohort (Mann-Whitney $U=132.00$ , $p<0.001$ , Cohen’s $d=1.29$ ). c–d, Comparison of individual functional connectivity correlations versus integrated model prediction for TI ( c ) and DBS ( d ). The first four panels show Spearman correlations between clinical improvement and connectivity measures from different brain regions (left STN-GPi, left STN-Thalamus, STN-M1, STN-SMA). The fifth panel displays the iVB-based model results pooled from a five-fold cross-validation test data, which achieves significantly stronger correlation with treatment outcomes compared to individual connectivity measures. e–f, Model prediction ROC, and Precision-Recall curves, prediction spectrum plots, for external ( e , n = 14) and prospective ( f , n = 11) validation. Curves show mean ± std over five random seeds, with BrainLM and BrainGNN as baselines.

Figure 5: Mechanistic validation of iVB-based model via prediction, attribution, and symptom dynamics. a-b, Pre-treatment iVB-derived regional CBM predicts TI response. a, Brain map visualizes AAL3 regions most strongly associated with TI response, colored by Cohen’s $d$ ; inset table lists the regions meeting the dual threshold ( $p<0.05$ , two-sided; $d>0.5$ ), reporting uncorrected $p$ , Cohen’s $d$ , and FDR-adjusted $p$ . b, Violin plots show CBM deviation distributions in these regions for responders versus non-responders. c, Shapley value analysis of the top 15 brain regions contributing to positive predictions in the TI cohort. d, Same Shapley value analysis for the DBS cohort. e–h, Region-symptom associations revealed by longitudinal iVB analysis. e, Heatmap showing Spearman correlations ( $|\rho|$ ) between changes in regional deviation (post–pre) derived from iVBs and improvements in MDS-UPDRS III motor scores for TI cohorts. Only associations with $|\rho|\geq 0.3$ and $p<0.05$ are displayed (non-significant values omitted). The panel shows a 6 $\times$ 6 submatrix of the six most frequently implicated brain regions and MDS-UPDRS III items. f, Representative scatter plot illustrating the strongest region-symptom association with Spearman correlation coefficient and $p$ -value indicated. g, h , Same analysis as e-f for the DBS cohort, where g is the heatmap that characterizes the correlation and h gives the representative scatter plot.

Extended Data Fig. 1 : Full Heatmap of absolute Spearman correlation strengths between brain regions and MDS-UPDRS symptom measures. Rows denote individual symptoms. Columns show the top-ranked regions, defined by the number of associations exceeding the threshold ( $|\rho|\geq 0.3$ and $p<0.05$ ). Color represents $|\rho|$ , with darker shades indicating stronger effects. Associations not meeting the threshold or not confirmed by the significance table are displayed in gray. Results are presented for TI and DBS cohorts separately.

Feature Type	AUC	AUPR	ACC	F1	Precision
Static_FC	0.417 ± 0.283	0.609 ± 0.167	0.600 ± 0.178	0.418 ± 0.378	0.450 ± 0.371
iVB_FC	0.420 ± 0.152	0.578 ± 0.115	0.582 ± 0.136	0.544 ± 0.276	0.568 ± 0.326
iVB_EC	0.784 ± 0.090	0.850 ± 0.076	0.727 ± 0.100	0.773 ± 0.074	0.781 ± 0.158
CBM	0.865 ± 0.072	0.915 ± 0.048	0.836 ± 0.036	0.847 ± 0.036	0.917 ± 0.105

Table 1: Ablation Study on CBM

AAL3 Label	Anatomical Name	AAL3 Label	Anatomical Name
Angular	Angular Gyrus	Putamen	Putamen
Caudate	Caudate Nucleus	Red_N	Red Nucleus
Cerebellum_3	Lobule III of Cerebellar Hemisphere	Rolandic_Oper	Rolandic Operculum
Cerebellum_4_5	Lobule IV, V of Cerebellar Hemisphere	Supp_Motor_Area	Supplementary Motor Area
Cerebellum_6	Lobule VI of Cerebellar Hemisphere	Temporal_Inf	Inferior Temporal Gyrus
Cerebellum_8	Lobule VIII of Cerebellar Hemisphere	Thal_IL	Intralaminar Thalamic Nucleus

Table 2: AAL3 Atlas Region Labels in Main Text and Corresponding Anatomical Names.

Dataset	n	Age (y)	Male n (%)	Race (W, B, A, I, H, O)	Label Distribution
HCP	812	N/A	402 (49.5)	887, 193, 64, 2, 5, 45	HC: 812
PPMI	579	65.9 ± 13.9	326 (56.3)	559, 4, 7, 3, 0, 6	HC: 406; PD: 173
ABIDE	1,097	17.0 ± 8.0	934 (85.1)	N/A	HC: 570; Autism: 527
ADNI	219	73.1 ± 7.2	101 (46.1)	197, 11, 4, 0, 0, 7	HC: 82; MCI: 103; AD: 34

Table 3: Public Dataset Study Population

Dataset	Source	n	Age (y)	Male n (%)	Target n (GPi/STN)
Retrospective TI	Ruijin Hospital (Shanghai)	51	69.84 ± 6.54	29 (56.9)	24/27
Retrospective DBS	Ruijin Hospital (Shanghai)	55	64.05 ± 8.27	37 (67.3)	28/27
Prospective TI	Ruijin Hospital (Shanghai)	11	60.27 ± 8.60	9 (81.8)	4/7
External TI	Zhongnan Hospital (Wuhan)	14	N/A	8 (57.1)	2/12

Table 4: Clinical Dataset Study Population

Clinical Outcome Prediction

Clinical outcome prediction uses patient records, imaging, signals, and biomarkers to estimate risks, disease trajectories, treatment response, and survival.

See all 265 papers

Treatment-Conditioned Diffusion for Forecasting Neurodegenerative Disease Progression
May 28, 2026Danylo Boiko, Viktoriia MishkurovaLongitudinal Medical Image Analysis Clinical Outcome Prediction
Innoloft Inc. 701 Brazos Street, Austin, TX 78701, USA · Bogomolets National Medical University Beresteiskyi Avenue, 34, Kyiv, 03057, Ukraine
PreSight: Preoperative Outcome Prediction for Parkinson's Disease via Region-Prior Morphometry and Patient-Specific Weighting
Mar 2, 2026Yand Wang, Chen Zhang, Lanyun Zhu +6Clinical Outcome Prediction Clinical Decision-Making
Beijing Jiaotong University · Nanyang Technological University · Institute of Medical Technology, Peking University +3
Brain-DiT: A Universal Multi-state fMRI Foundation Model with Metadata-Conditioned Pretraining
Apr 14, 2026Junfeng Xia, Wenhao Ye, Xuanye Pan +3Medical Imaging Foundation Models Diffusion Transformer
Department of Biomedical Engineering, Southern University of Science and Technology, China · School of Biomedical Engineering, Shenzhen University, China

May 28, 2026J/K move · Enter open · S save

Treatment-Conditioned Diffusion for Forecasting Neurodegenerative Disease Progression

Danylo Boiko, Viktoriia Mishkurova

Innoloft Inc. 701 Brazos Street, Austin, TX 78701, USA · Bogomolets National Medical University Beresteiskyi Avenue, 34, Kyiv, 03057, Ukraine

Longitudinal Medical Image Analysis Clinical Outcome Prediction Medical Image Synthesis Parkinson's Disease Conditional Diffusion Models

Open paper AskPDF arXiv

Forecasting the progression of neurodegenerative diseases, such as Parkinson's disease, is essential for effective long-term planning and personalized therapeutic intervention. Existing systems typically produce scalar clinical scores that ignore the rich structure of longitudinal neuroimaging, while traditional generative approaches suffer from a loss of anatomical details and blurring subtle progression patterns. To address this, we introduce a novel treatment-conditioned diffusion framework that predicts high-fidelity future brain states by conditioning the generative process on patients' screening DaTscan images and levodopa equivalent daily dose over one year. The pipeline uses a Transformer-based encoder to represent non-linear, time-dependent pharmacological dynamics and optimizes generation through a multi-weight region-of-interest mask that focuses on biologically critical areas. Experimental evaluation shows that our framework maintains sharp anatomical boundaries and significantly improves clinical fidelity relative to the baseline, achieving 14.0% lower MSE, 7.2% lower MAE, and 4.9% higher SSIM.

Abstract

Clinical Outcome Prediction

Explore similar work

Treatment-Conditioned Diffusion for Forecasting Neurodegenerative Disease Progression

PreSight: Preoperative Outcome Prediction for Parkinson's Disease via Region-Prior Morphometry and Patient-Specific Weighting

Brain-DiT: A Universal Multi-state fMRI Foundation Model with Metadata-Conditioned Pretraining