Deep learning applied to deep brain stimulation in parkinson’s disease

Abstract

In order to better model complex real-world data such as biomedical signals, one approach is to develop pattern recognition techniques and robust features that capture the relevant information. In this paper, we use deep learning methods, and in particular multilayer perceptron, to build an algorithm that can predict subcortical structures of patients with Parkinson’s disease, based on microelectrode records obtained during deep brain stimulation. We report on experiments using a data set involving 52 microelectrode records for the structures: zona incerta, subthalamic nucleus, thalamus nucleus, and substantia nigra. The results show that the combination of features and deep learning produces 99.2% precision of detection and classification on the average of the subcortical structures under study. In conclusion, based on the high precision obtained in the classification, deep learning could be used to predict subcortical structure, and mainly the subthalamic nucleus for neurostimulation.