Investigating ResNet deep features for Parkinson’s disease diagnosis using hand-drawn pattern

Parkinson’s is one of the most common diseases in which the patient suffers from a disorder involving shaking and improper muscle balance and coordination. This makes their daily life activities quite different and troublesome from healthy normal individuals. This paper deals with the detection of patients afflicted with Parkinson’s disease and a normal healthy person based on a dataset that involves hand-drawn spiral and wave structures by them. After the image processing of these hand-drawn structures, a deep learning algorithmic approach is implemented to detect how accurately a model can predict whether the drawing would be made by a healthy person or a person suffering from Parkinson’s disease. The model incorporated here is Resnet-50 architecture having enhanced performance owing to the large number of layers used and has a higher speed. The results were obtained over a range of iterations performed using this model concerning several parameters. Significant and accurate predictions for the disease detection were achieved therefore making this approach more effective to be implemented while using more complicated datasets with larger deep learning architectures.