An Early Detection of Parkinson’s Disease from Geometric Drawings

Abstract

Parkinson’s disease (PD) is a sensory system issue that may prompt shaking, firmness and trouble in strolling. Hence, it is a non-communicable disease; hence, proper diagnosis will prevent further damage to the body at an early stage. Most of the symptoms occur due to a decrease in the dopamine level in the patient’s body. Literature has shown that it is feasible to distinguish PD by requesting that the patient draw a spiral or wave and track their drawing and pen pressure speed afterward. The drawing speed is increasingly slow pen pressure is lower for an individual having the PD. The methodology uses the Histogram of Oriented Gradients (HOG) as the feature descriptor and proposed the weighted Random Forest (WRF) classifier technique. HOG will track the intensity changes in the images, and the WRF works with a small dataset to provide effective results. This strategy provides a very good testing accuracy of 93% and 92% on the wave and spiral datasets. This method is a very robust and cost-effective method for the early detection of PD.