Long-term local field potentials (LFPs) were recorded from eight STNs (four PD patients) during 92 recording sessions (44 OFF and 48 ON levodopa states), over a period of 3-12 months. Electrophysiological analysis included the power of frequency bands, band power ratio and burst features. A total of 140 engineered features was extracted for 20,040 epochs (each epoch lasting 5 seconds). Based on these engineered features, machine-learning models classified LFPs as OFF vs ON levodopa states.
Beta and gamma band activity alone poorly predicts OFF vs ON levodopa states, with an accuracy of 0.66 and 0.64, respectively. Group machine-learning analysis slightly improved prediction rates, but personalized machine-learning analysis, based on individualized engineered electrophysiological features, were markedly better, predicting OFF vs ON levodopa states with an accuracy of 0.8 for support vector machine learning models.
We showed that individual patients have unique sets of STN neurophysiological biomarkers that can be detected over long periods of time. Machine-learning models revealed that personally classified engineered features most accurately predict OFF vs ON levodopa states. Future development of aDBS for PD patients might include personalized machine-learning algorithms. Clinical trials: This study was supervised and authorized by the Ethics Committees of the IRB of the Hadassah Medical Center (no.0403-13-HMO) and the Israel Ministry of Health (no.HT6752), and it received clinical trials registration (no. NCT01962194).
© 2021 IOP Publishing Ltd.