Parkinson’s Disease Classification from Magnetic Resonance Images (MRI) using Deep Transfer Learned Convolutional Neural Networks

Abstract

Parkinson’s Disease (PD) is a progressive brain disorder cased by dopmainergic neuronal loss and mainly affects the Substantia Nigra located in the mid brain region. The increasing availability of public datasets has driven the development of advanced machine learning algorithms as a tool to assist in the classification and initial risk assessment of patients with PD. This work provides an analysis of five major deep learning architectures with the aim of refinement of Magnetic Resonance Imaging (MRI) based diagnosis of PD, evaluated using multiple performance indices. Three of the five architectures considered show a significant increase in performance in comparison to existing work without hyper-parameter tuning and can aid researchers in selecting a Deep Neural Network (DNN) model as an MRI based classification model for PD. The results support and demonstrate the scope for the use of Artificial Intelligence (AI) as a decision support system.