A Transfer Learning Approach with MobileNetV2 for Parkinson’s Disease Detection using Hand-Drawings
B. Anil Kumar, Mohan Bansal
Abstract
Parkinson’s disease is a condition that affects common human movements because of the brain’s failing neurons. The primary cause of this disease is a deficiency of dopamine in the brain, which can also cause changes in blood pressure, eating difficulties, and disrupted sleep. For patients to receive proper treatment and to improve their health, it is essential to detect Parkinson’s disease at an early stage, as there is currently no cure for the disease. To address this, our research focuses on using spiral and wave hand-drawings as a means of early detection. We developed a modified MobileNetV2 approach with deep learning to accurately predict Parkinson’s disease using these drawings. Our approach achieved a high accuracy of 97.70%, with a low error rate of 2.30%, while using fewer parameters than other models. Our findings suggest that using hand-drawings as a diagnostic tool can greatly improve the accuracy of Parkinson’s disease diagnosis.