Dokument: From Neural Oscillations to Clinical Decisions: Machine Learning for Predicting Optimal Deep Brain Stimulation Parameters in Parkinson's Disease
| Titel: | From Neural Oscillations to Clinical Decisions: Machine Learning for Predicting Optimal Deep Brain Stimulation Parameters in Parkinson's Disease | |||||||
| Weiterer Titel: | - | |||||||
| URL für Lesezeichen: | https://docserv.uni-duesseldorf.de/servlets/DocumentServlet?id=73251 | |||||||
| URN (NBN): | urn:nbn:de:hbz:061-20260513-150525-8 | |||||||
| Kollektion: | Dissertationen | |||||||
| Sprache: | Englisch | |||||||
| Dokumententyp: | Wissenschaftliche Abschlussarbeiten » Dissertation | |||||||
| Medientyp: | Text | |||||||
| Autor: | Rassoulou, Fayed [Autor] | |||||||
| Dateien: |
| |||||||
| Beitragende: | Prof. Dr. Schnitzler, Alfons [Gutachter] Zimmermann, Eckart [Gutachter] | |||||||
| Stichwörter: | Parkinson Disease / Deep Brain Stimulation | |||||||
| Dewey Dezimal-Klassifikation: | 600 Technik, Medizin, angewandte Wissenschaften » 610 Medizin und Gesundheit | |||||||
| Beschreibungen: | Deep brain stimulation (DBS) of the subthalamic nucleus (STN) provides effective symptom relief for patients with advanced Parkinson's disease (PD). However, DBS programming remains challenging: identifying optimal electrode contacts currently relies on time-intensive monopolar reviews where neurologists systematically test all contacts through trial-and-error. This process lacks objective biomarkers and depends heavily on clinical expertise. Decades of research have characterized pathological oscillatory signatures in PD, yet these neural biomarkers have remained largely confined to basic research without informing clinical decision-making.
This thesis investigated whether electrophysiological recordings can predict therapeutic outcomes and guide contact selection. I pursued three aims: first, curating a unique dataset of simultaneous magnetoencephalography and local field potential recordings from twenty PD patients, organized according to Brain Imaging Data Structure (BIDS) standards; second, applying machine learning to this dataset to identify which neural signatures predict the therapeutic window of individual electrode contacts; third, validating predictions in an independent cohort of eight newly recruited patients. Machine learning successfully predicted therapeutic windows based on neurophysiological markers, achieving significant correlations in both the training cohort (r = 0.45, p < 0.001) and independent validation cohort (r = 0.30, p < 0.001). Predictions relied mainly on high-frequency subthalamic activity (>35 Hz) and STN-cortex coherence patterns across multiple brain regions, rather than the widely studied beta oscillations. This dissociation likely reflects different functional roles: beta power indicates disease severity, while high-frequency signals and connectivity patterns better mark optimal electrode for stimulation. This work establishes proof-of-concept that electrophysiological signatures contain information about clinically relevant parameters, such as the optimal DBS contact. However, important limitations must be acknowledged: magnetoencephalography (MEG) is not widely available; the therapeutic window does not capture long-term quality of life; and sample sizes were relatively small, though successful validation in an independent cohort mitigates overfitting concerns. Critically, this thesis demonstrates feasibility rather than clinical implementation. Whether electrophysiology-guided programming improves outcomes compared to standard approaches would require prospective randomized trials. Nonetheless, by demonstrating that prediction is possible, this work provides a foundation for future translational efforts and contributes to evolving DBS therapy toward data-driven precision medicine.- | |||||||
| Lizenz: |  Dieses Werk ist lizenziert unter einer Creative Commons Namensnennung 4.0 International Lizenz | |||||||
| Fachbereich / Einrichtung: | Mathematisch- Naturwissenschaftliche Fakultät | |||||||
| Dokument erstellt am: | 13.05.2026 | |||||||
| Dateien geändert am: | 13.05.2026 | |||||||
| Promotionsantrag am: | 03.03.2026 | |||||||
| Datum der Promotion: | 07.05.2026 |
