Detection Of Parkinson’s Disease And Its Severity Using Deep Learning

Abstract

Parkinson’s disease (PD) is a chronic and progressive neurological disorder that affects millions of individuals worldwide. The condition is primarily characterized by the degeneration of dopamine-producing neurons, which leads to impairments in motor control, tremors, rigidity, and slowed movement. Detecting Parkinson’s disease in its early stages remains difficult because there is no single definitive clinical diagnostic test. As a result, many patients are diagnosed only after significant neurological damage has already occurred. Early identification is therefore essential to improve treatment outcomes, reduce disease progression, and enhance patient quality of life.

Recent clinical studies indicate that handwriting abnormalities are closely associated with Parkinson’s disease. Individuals with PD often display symptoms such as tremor-induced distortions, reduced writing speed, irregular stroke patterns, inconsistent pressure, and impaired fine motor control. Among various handwriting assessments, spiral and wave drawing tests are widely used in neurological evaluation because they effectively capture subtle motor impairments. The distortions and irregularities present in these patterns can serve as important indicators for early PD detection.This research proposes an automated system that analyzes spiral and wave sketches to detect and evaluate Parkinson’s disease severity. The framework employs two dedicated Convolutional Neural Network (CNN) models to independently process spiral and wave images. These deep learning models automatically learn spatial features such as tremor frequency, stroke instability, irregular curvature, and drawing distortion. The system is implemented using the TensorFlow deep learning framework in Python.