Unsupervised Clustering of Micro-Electrophysiological Signals for Localization of Subthalamic Nucleus During DBS Surgery

Abstract

In this paper, an unsupervised machine learning technique is proposed to localize Subthalamic Nucleus (STN) during deep brain stimulation (DBS) Surgery. DBS is one of most common treatments for advanced Parkinson's disease (PD). The purpose of this surgery is to permanently implant stimulation electrodes inside the STN to deliver electrical currents. It is clinically shown that DBS surgery can significantly reduce motor symptoms of PD (such as tremor). However, the outcome of this surgery is highly dependent on the location of the stimulating electrode. Since STN is a very small region inside the basal ganglia, accurate placement of the electrode is a challenging task for the surgical team. During DBS surgery, the team uses Micro-Electrode Recording (MER) of electrophysiological neural activities to intraoperatively track the location of electrodes and estimate the borders of the STN. In this work, we propose a composite unsupervised machine learning clustering approach that is capable of detecting the dorsal borders of the STN during DBS operation. For this, MER signals from 50 PD patients were recorded and used to validate the performance of the proposed method. Results show that the approach is capable of detecting the dorsal border of STN in an online manner with an accuracy of 80% without using any supervised training.