Parkinson’s disease based on deep learning using MR images

Abstract

Magnetic resonance imaging has become an indispensable aid in the Parkinson's disease. The traditional method of analysis is that the patient takes an MR image of the brain and then the doctor analyzes the MR image for diagnosis. However, due to the poor quality of MR images and the high noise, it is difficult for doctors to diagnose, and the professional requirements for doctors are relatively high when analyzing nuclear magnetic resonance images. This paper proposed a MR image classification algorithm based on deep learning. The algorithm is mainly divided into two phases. The first phase is to detect the diagnostic region of the MR image by the improved Faster RCNN network. The second phase is to classify the diagnostic areas detected in the previous phase through our custom CNN network. We tested the algorithm using MR images of Parkinson's disease. The experimental results show that the accuracy of MR image detection and classification can be greatly improved by algorithm improvement.