Electrophysiological Markers Predict Optimal DBS Electrode Contacts in Parkinson’s Disease

Overview

This study demonstrates that electrophysiological features from subthalamic nucleus (STN) recordings and STN-cortex coherence can predict the therapeutic window of deep brain stimulation (DBS) contacts in Parkinson’s disease. Using machine learning on resting-state MEG and local field potential data, the model reliably identified optimal contacts, potentially streamlining DBS programming.

Background

Deep brain stimulation targeting the subthalamic nucleus is an established treatment for Parkinson’s disease, aiming to alleviate motor symptoms by modulating pathological neural oscillations. Programming DBS devices to find the optimal electrode contact is challenging due to the complex anatomy and risk of side effects from non-target stimulation. Current clinical practice involves time-consuming monopolar reviews to identify contacts with the largest therapeutic window. Machine learning approaches leveraging electrophysiological signals offer a promising avenue to automate and accelerate this process.

Data Highlights

Key Findings

• Machine learning models using STN power and STN-cortex coherence predict the therapeutic window of DBS contacts with significant accuracy.
• Fast subthalamic activity (>35 Hz) and coherence in multiple frequency bands were the most informative electrophysiological features.
• The model generalized well to an independent patient cohort, confirming robustness.
• Predicted contact rankings allow faster identification of the optimal contact, potentially reducing programming time.
• Combining multiple electrophysiological markers provides a more comprehensive approach than relying on single signals like beta power alone.

Clinical Implications

Incorporating electrophysiological markers into DBS programming can facilitate automated and efficient selection of optimal electrode contacts, reducing the burden of monopolar reviews. This approach may improve clinical outcomes by maximizing the therapeutic window and minimizing side effects. Future integration into clinical workflows could enhance personalized DBS therapy.

Conclusion

The study establishes the feasibility of using electrophysiological features combined with machine learning to predict optimal DBS electrode contacts, offering a promising tool to support and accelerate clinical programming in Parkinson’s disease.

References

1. Roediger et al. 2021 -- Automated Programming of Deep Brain Stimulation
2. Sarikhani et al. 2022 -- Optimizing DBS Programming with Smartwatch Accelerometry
3. Original Study -- Electrophysiological Markers Indicate the Optimal Therapeutic Range for Deep Brain Stimulation Electrode Contacts