Objective: The objective of this study was to develop a machine learning (ML) algorithm capable of predicting motor phenotypes in Parkinson’s disease (PD) using intraoperative neurophysiology.

Background: Specific spectral features from local field potential (LFP) recordings in the pallidum and the subthalamic nucleus (STN) have been shown to correlate with motor symptoms in PD. More specifically, elevated beta band (13-30 Hz) power has been linked to the severity of motor symptoms such as bradykinesia and rigidity [1], while tremor has been associated with suppression of beta band power [2] , increase in gamma band (35–55 Hz) power [3] and theta band (4.5 -5.5 Hz) power [4]. However, the combined utility of these neurophysiological markers for delineating PD motor phenotypes has not been explored.

Method: We conducted a retrospective study of patients who underwent pallidal or STN DBS with intraoperative resting-state local field potential recordings from all monopolar contacts on the DBS lead. The power spectral density (PSD) was computed and normalized across patients. Patients were categorized into tremor and non-tremor phenotypes based on their preoperative Unified Parkinson’s Disease Rating Scale scores [5]. Various ML models were then iteratively constructed and evaluated, performing feature selection to isolate the most relevant frequencies. The ML models were evaluated by 5-fold cross-validation and a holdout test set. Our primary performance metric was the area under receiver operating characteristic curve (AUC-ROC).

Results: A total of 137 DBS implantations (94 GPi, 43 STN) representing 110 unique patients were included in the study cohort. Our best model was able to identify the correct motor phenotype with 72% accuracy with an overall AUROC of 74%. The features selected in the model included PSDs in the alpha, beta, and gamma bands.

Conclusion: Spectral activity from LFP recordings shows promise to predict motor phenotypes in PD. Our findings lay the foundation for developing an automated algorithm as an objective measure of PD phenomenology that could be critical for emerging adaptive DBS paradigms.

AUC ROC curve of Decision Trees model.

References: [1] Neumann, W.-J., Staub-Bartelt, F., Horn, A., Schanda, J., Schneider, G.-H., Brown, P., & Kühn, A. A. (2017). Long term correlation of subthalamic beta band activity with motor impairment in patients with Parkinson’s disease. Clinical Neurophysiology, 128(11), 2286–2291. https://doi.org/10.1016/j.clinph.2017.08.028.

[2] Qasim, S. E., de Hemptinne, C., Swann, N. C., Miocinovic, S., Ostrem, J. L., & Starr, P. A. (2016). Electrocorticography reveals beta desynchronization in the basal ganglia-cortical loop during rest tremor in Parkinson’s disease. Neurobiology of Disease, 86, 177–186. https://doi.org/10.1016/j.nbd.2015.11.023 .

[3] Weinberger, M., Hutchison, W. D., Lozano, A. M., Hodaie, M., & Dostrovsky, J. O. (2009). Increased Gamma Oscillatory Activity in the Subthalamic Nucleus During Tremor in Parkinson’s Disease Patients. Journal of Neurophysiology, 101(2), 789–802. https://doi.org/10.1152/jn.90837.2008

[4] Reck, C., Florin, E., Wojtecki, L., Krause, H., Groiss, S., Voges, J., Maarouf, M., Sturm, V., Schnitzler, A., & Timmermann, L. (2009). Characterisation of tremor-associated local field potentials in the subthalamic nucleus in Parkinson’s disease. European Journal of Neuroscience, 29(3), 599–612. https://doi.org/10.1111/j.1460-9568.2008.06597.x

[5] Stebbins, G. T., Goetz, C. G., Burn, D. J., Jankovic, J., Khoo, T. K., & Tilley, B. C. (2013). How to identify tremor dominant and postural instability/gait difficulty groups with the movement disorder society unified Parkinson’s disease rating scale: Comparison with the unified Parkinson’s disease rating scale. Movement Disorders, 28(5), 668–670. https://doi.org/10.1002/mds.25383

« Back to 2025 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/parkinsons-disease-motor-phenotypes-delineated-by-pallidal-and-subthalamic-neurophysiology-and-machine-learning/