Objective: To develop a machine learning (ML) algorithm capable of predicting the optimal deep brain stimulation (DBS) contact based on neurophysiology.

Background: DBS is an effective therapy for medication-resistant Parkinson’s disease (PD), but individual patient outcomes can vary significantly. Currently, optimization of stimulation parameters is a manual and time-consuming process based on subjective assessments that can take weeks to months [1]. Recent advancements in DBS technology permit recording of neural physiology in the form of local field potentials (LFPs). In PD, LFP spectral features such as beta band (13-30Hz) correlate with motor symptom severity [2]. However, whether LFP can be used to predict optimal stimulation parameters is still unknown.

Method: We conducted a retrospective study with data collected from PD patients implanted with DBS leads in either the subthalamic nucleus (STN) or the globus pallidus internus (GPi), and paired with a DBS system capable of LFP recordings (Percept, Medtronic, Minneapolis, MN). The LFPs were recorded from 3 bipolar contacts in the DBS OFF state between 1 week to 3 months after the device implantation. For each patient, the therapeutic stimulation parameters determined after 6 months of DBS programming were collected and used as the ‘optimal settings’. The power spectral density (PSD) was computed and used to construct various ML models predicting the optimized contact. The ML model training paradigm included 5-fold cross-validation and evaluated on a holdout test. An ensemble model was also developed based on thresholding PSDs into subsets by signal variability. For comparison, predicting the optimal contact based on only maximum or average beta power was also calculated.

Results: Recordings from 98 brain hemispheres (86 GPi, 12 STN) [KJ1] from 72 patients were included in the analysis. Our preliminary ML model was able to predict the optimal contact with a balanced accuracy of 60%. However, our ensemble model based on thresholding optimization achieved a balanced accuracy of 90%. The ML models were more accurate at predicting the optimal contact than using the maximum or average beta power (47% accuracy).

Conclusion: The findings indicate that LFP can be used to guide the selection of therapeutic DBS contacts. This lays the groundwork for developing an automated algorithm that can substantially reduce the clinical workload and duration of DBS programming.

Signal processing workflow

References: [1] Patel, B.; Chiu, S.; Wong, J. K.; Patterson, A.; Deeb, W.; Burns, M.; Zeilman, P.; Wagle-Shukla, A.; Almeida, L.; Okun, M. S.; and Ramirez-Zamora, A. 2021. Deep brain stimulation programming strategies: segmented leads, independent current sources, and future technology. Expert Review of Medical Devices, 18(9): 875–891. Publisher: Taylor & Francis eprint: https://doi.org/10.1080/17434440.2021.1962286.

[2] Van Wijk, B. C. M., De Bie, R. M. A., & Beudel, M. (2023). A systematic review of local field potential physiomarkers in Parkinson’s disease: From clinical correlations to adaptive deep brain stimulation algorithms. Journal of Neurology, 270(2), 1162–1177. https://doi.org/10.1007/s00415-022-11388-1.

« Back to 2025 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/towards-automated-deep-brain-stimulation-programming-using-neurophysiology-and-artificial-intelligence/