Parkinson’s Disease Diagnosis Using Deep Convolutional Neural Networks

Parkinson’s disease (PD), is a type of neurodegenerative disease, in the initial stages it targets the human brain and this is a slow process. The symptoms for this disease are constraint in muscles, shaking of limbs, and unusual balance, with time the given symptoms start getting worse. Within this paper, we have investigated and experimented with the usage of a VGG-19 model which is fine-tuned for diagnosing Parkinson’s Disease (PD) based on real-world images drawn by Parkinson’s Disease patients. Pre-processing of 102 wave & 102 spiral drawings from one of the sets of images and 368 wave & 368 meander drawings from the other set of images was done. The pre-processing included the scaling of images along with data augmentation to enhance the precision of the CNN. Training of the Convolutional Neural Network (CNN) model was done on the images that underwent prior treatment. And validation was done with the help of the validation images. The accuracy of this model was determined to be 88% and 89% for the images in Set-1 and Set-2 respectively.