Deep Learning Application for Monitoring Handwritten Spirals in Parkinson's Disease

Parkinson’s disease (PD) is a prevalent neurodegenerative disorder that affects the central nervous system and progressively impairs the control of voluntary movements. The growing body of published research in this field demonstrates the increasing attention that spiral analysis (SA) has attracted in studies focused on the diagnosis and monitoring of PD. SA-based detection methods have shown superior performance compared to traditional approaches involving clinical evaluations, laboratory tests, and neuroimaging, largely due to their practicality and cost-efficiency. This study investigates the potential of spiral drawing tasks as a tool for monitoring Parkinson’s disease by applying machine learning (ML) techniques to assess disease status. Specifically, four convolutional neural network (CNN) architectures—VGG16, ResNet50, MobileNetV2, and Xception—were employed to classify individuals into two groups—Parkinson’s Disease (PD) and Control—and to detect disease stages within the PD group based on the HoehnandYahrscaleacross multiple data collection sessions. Classification performance was assessed using accuracy and F1-score metrics. The results demonstrate that spiral-based assessments, combined with ML approaches, can effectively support both the diagnosis and longitudinal monitoring of Parkinson’s disease progression.