Objective: To assess the predictive value of functional and volumetric MRI data in comparison to clinical data alone for STN DBS outcomes.

Background: While STN DBS is effective for the treatment of Parkinson disease (PD), patient selection remains challenging due to the inadequacy of traditional levodopa response as a predictor of outcomes^1,2. Several MRI-based predictors of STN DBS outcome have previously been identified, including resting state functional connectivity and volumetric measures^3,4. We sought to predict DBS outcomes and compare predictive performance using clinical and MRI-based measures, using our most recent methods for data processing, model building, and analysis.

Method: We analyzed 65 participants who received STN DBS for PD at Washington University between 2007 and 2017. We used relaxed lasso (penalized, MATLAB 2022a) to select factors and generate models predicting percent change improvement in motor scores over 12 months following STN DBS. Multiple predictor sets were used to compare performance of clinical, volumetric, and FC variables, with leave-one-out cross-validation to evaluate out-of-sample model performance with R² and root mean squared error (RMSE). We used standardized regression coefficients (β) for the optimized model to compare relative influence of individual predictors.

Results: Addition of FC and volumetric predictors improved cross-validated model performance over clinical predictors alone (clinical: R² = 0.1490, RMSE 14.22; imaging + clinical: R² = 0.3153, RMSE 12.59). For the optimized model with all available predictors, the most influential predictors included 1) percent levodopa response (β = 0.6156), 2) Levodopa daily equivalent dose (β = 0.3389), 3) FC components including basal ganglia and ventral attention networks (β = 0.3058), and 4) FC components including cerebellar and salience networks (β = -0.2154).

Conclusion: MRI-based predictors improve model performance for prediction of STN DBS motor outcomes over clinical predictors alone, with FC predictors most influential in an optimized model. Potential bias exists here due to levodopa response being used to select these patients for DBS initially. Validation of these predictors in an external cohort will be necessary to confirm their utility in routine clinical practice.

References: 1. Wolke R, Becktepe JS, Paschen S, et al. The Role of Levodopa Challenge in Predicting the Outcome of Subthalamic Deep Brain Stimulation. Mov Disord Clin Pract 2023;10(8):1181–1191.
2. Zaidel A, Bergman H, Ritov Y, Israel Z. Levodopa and subthalamic deep brain stimulation responses are not congruent. [Internet]. Mov Disord 2010;25(14):2379–86.Available from: http://www.ncbi.nlm.nih.gov/pubmed/20824733
3. Younce JR, Campbell MC, Perlmutter JS, Norris SA. Thalamic and ventricular volumes predict motor response to deep brain stimulation for Parkinson’s disease. [Internet]. Parkinsonism Relat Disord 2019;61(November):64–69.Available from: https://linkinghub.elsevier.com/retrieve/pii/S1353802018305236
4. Younce JR, Campbell MC, Hershey T, et al. Resting-State Functional Connectivity Predicts STN DBS Clinical Response. [Internet]. Mov Disord 2020;Available from: http://www.ncbi.nlm.nih.gov/pubmed/33211330

« Back to 2024 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/functional-connectivity-enhances-prediction-of-stn-dbs-motor-outcomes/