Parkinson Detection using Image Convolutional Neural Network and Transfer Learning

Abstract

Parkinson's disease is a progressive disease of the nervous system. The disorder affects several regions of the brain, most notably an area called the substantia nigra which controls balance and movement. Often the first symptom of Parkinson's disease is tremors or tremors in a limb, especially when the body is at rest. The tremor usually begins on one side of the body, usually in one hand. Tremors can also affect the arms, legs, feet and face. Stiffness in limbs and chest, slow movement means bradykinesia and postural instability are other characteristic symptoms of Parkinson. These symptoms slowly worsen over time. In this article, we present a Parkinson's detection system. To train, model and predict Parkinson's, transfer using the VGG-16 network. The VGG-16 network trained hand-drawn images of spirals and waves. In this system, the user can submit their image in a hand-drawn fashion and send it to the application which will train that image in the model and provide a prediction on Parkinson's. If the user has Parkinson's, this system can be used to provide medical and therapeutic information for further diagnosis. The evaluation is made on the ability to detect different directions during drawing movements, obtaining the best results for the X and Y directions. The experimental results show that the system provides an accuracy of 96.5%.