Prediction of STN-DBS outcome in Parkinson’s disease using machine learning

This is the authors' abstract. We don't add key points for this paper.

Deep brain stimulation (DBS) targeting the subthalamic nucleus (STN) is an established therapy for advanced Parkinson's disease (PD), but outcomes vary significantly among patients.Using a dataset of 408-420 PD patients (depending on outcome), we developed machine learning models to predict outcomes of STN-DBS based on preoperative clinical markers.Regression models predicted scores on the Movement Disorders Society Unified Parkinson's Disease Rating Scale (MDS-UPDRS) part III, and subscores for Tremor, Axial symptoms, and Bradykinesia & Rigidity.The models achieved root mean square errors (RMSE) of 9.1, 2.6, 2.5, and 5.3, respectively.These results demonstrate the models' ability to provide accurate predictions despite the heterogeneity of PD.This approach refines patient selection by forecasting postoperative outcomes and enables personalized treatment planning.Future iterations will explore additional predictors, such as neuroimaging data, to further improve model performance and support clinical decision-making in DBS therapy.This study advances the use of machine learning in predictive medicine for PD.