Machine learning Approaches for Detecting Dorsolateral Oscillatory Region (DLOR)

The detection process of Subthalamic Nucleus (STN) and its nearer surroundings as Dorsolateral Oscillatory Region (DLOR) is currently carried out by human experts, with the help of supervised machine learning detection algorithms. So, the success of the process relies on the human experts that involved in effective labelled data used to train various machine learning algorithms. Thus, we have considered an unsupervised algorithm to identify the STN and the DLOR during deep brain stimulation (DBS) surgery as per an agnostic modelling approach. Based on the provided data, we derived fresh features and calculate a modification of the Mahalanobis distance between them. Theoretically, we demonstrate that greater separation increases the differences among experiments that have different inherent characteristics. The manifold learning method based Diffusion Maps is enhanced by the addition of the new features and distance for providing measurements. We have considered different set of experiments which are collected from different patients at the time of surgery and produce detection performance. Our unsupervised approach achieves results that are on par with those of a supervised algorithm when it comes to identifying the STN and even better when it comes to identifying the DLOR.