Automatic identification of DBS parameters from the volume of tissue activated (VTA) using support vector machines

Abstract

Deep brain stimulation (DBS) is a neurosurgical method to treat symptoms of Parkinson’ disease. Several computational models, mostly based on finite element method (FEM) have been employed to describe the interaction of electromagnetic waves in brain tissues during DBS. Also, for planning the DBS, it is necessary to estimate with precision the neural response generated by electrodes in the stimulated region, what it is known as volume of tissue activated (VTA). However, this estimation should consider the intrinsic properties of each patient, therefore DBS parameters must be adjusted individually. In this work, we propose a 3D interaction module for estimating the DBS parameters (amplitude, contacts, among others) from a desired VTA using support vector machines (inverse problem). Also, we developed an interactive application for analyzing the VTA generated by DBS in the subthalamic nucleus (STN) combining medical imaging and non-rigid deformation models. This module is a part of the NEURONAV software, previously developed for clinical support during postoperative therapy of neuro-modulation performed in Colombian PD patients. Outcomes show that it is possible to estimate with high accuracy the DBS parameters for different subjects.