Towards Automated Stimulation Parameter Titration for Deep Brain Stimulation: A Connectivity-based Approach
Loading...
Date
2018-05
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
The Ohio State University
Abstract
Deep brain stimulation (DBS) is an effective surgical treatment for drug-resistant neurological movement disorders like Parkinson's disease and essential tremor. However, as the complexity of DBS leads increases, the standard method for stimulation parameter titration will become increasingly challenging and definitive parameter search more time-consuming. For this reason, we proposed a connectivity-based approach using patient-specific cortical connectivity for predicting clinical outcome, and designed our study to serve as a proof-of-concept for automating stimulation parameter titration. Specifically, we obtained brain images from a cohort of 24 Parkinson's patients implanted with subthalamic nucleus (STN) DBS and calculated the differences in connectivity to cortical regions in the whole brain associated with specific clinical observations sorted by clinical outcome: "Improvement" or "Side Effect." We then reduced the number of cortical regions using reverse sequential feature selection before training a linear support vector machine (SVM) using 10-fold cross validation to classify clinical observations by clinical outcome using their associated differences in cortical connectivity. This SVM was then used to predict the most efficacious contact for each DBS lead, as well as the most efficacious voltage at each contact on each lead. The SVM achieved an overall classification error of 10.99%, and predicted the most efficacious contact with an average absolute deviation of 0.89, 1.17 and 1.08 contacts and 0.77, 1.10 and 1.21 V from the initial, 1 year and final stimulation parameter settings respectively. Additionally, it was found that 44.1% of the contacts were non-efficacious with 6 leads being entirely non-efficacious. Though these results are relatively modest, together they suggest that our metric is informative for predicting clinical outcome and narrowing down stimulation parameters. In the future, we hope to refine our algorithm to improve performance considerably. Our study serves as a useful first step towards the automation of stimulation parameter titration for DBS surgery.
Description
Keywords
deep brain stimulation, machine learning, support vector machine, tractography, automation, modeling and simulation, whole brain connectivity, neuroanatomy, medicine, DBS, SVM