Online prediction of optimal deep brain stimulation contacts from local field potentials in chronically-implanted patients with Parkinson’s disease
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
The selection of optimal contacts for chronic deep brain stimulation (DBS) requires manual iterative testing of multiple stimulation configurations: the monopolar review. This requires time, highly trained personnel, and can cause patient discomfort. The use of neural biomarkers may help speed up this process.
Objective
This study aimed to validate the use of local field potentials (LFP) from a chronically implanted DBS neurostimulator to inform clinical selection of optimal stimulation contact-levels.
Methods
We retrospectively analysed bipolar LFP-recordings performed in patients with Parkinson’s disease OFF-medication and OFF-stimulation across three centres. For each contact-level chosen clinically, we ranked the recordings obtained by different channels according to the informative value of various beta-band (13-35Hz) power measures. We then developed two prediction algorithms: (i) a “decision-tree” method for direct, in-clinic use, and (ii) a “pattern based” method for offline validation. We finally compared these approaches to existing prediction algorithms.
Results
We included 68 subthalamic nuclei from the Netherlands (NL), 21 from Switzerland (CH), and 32 from Germany (DE). Recording channel rankings depended on the clinically chosen contact-level. When predicting the first two contact-levels, the online “decision tree” method achieved a predictive accuracy of 86.5% (NL), 86.7% (CH), and 75.0% (DE), respectively. The offline “pattern based” technique attained similar results. Both prediction techniques outperformed an existing algorithm and were robust in different clinical and recording conditions.
Conclusion
This study demonstrates that using these new methods, LFP-signals recorded in-clinic can support the selection of stimulation contact-levels, with high accuracy, reducing DBS programming time by half.
