Parkinson’s Disease Detection Using Spiral Drawing Analysis with 12-Feature Support Vector Machine Classification in MATLAB

Abstract

Parkinson’s disease (PD) is a progressive neurological disorder that affects motor control, hand movement, coordination, and fine motor skills. Spiral drawing analysis provides a non-invasive image-based approach for studying variations in motor control. This study presents a computer-aided Parkinson’s disease detection system based on spiral drawing images and Support Vector Machine (SVM) classification implemented in MATLAB. The dataset used in the experiment contained 528 spiral images, comprising 280 healthy samples and 248 Parkinson’s disease samples. Twelve handcrafted image features were extracted from each drawing: Area, Perimeter, Eccentricity, Circularity, Solidity, Extent, Major Axis Length, Minor Axis Length, Mean Intensity, Standard Deviation, Entropy, and Edge Density. The images were converted to grayscale, filtered, thresholded, cleaned using morphological operations, and reduced to the largest connected component before feature extraction. The feature vectors were standardized and classified using an RBF-kernel SVM. Hyperparameter optimization was performed using BoxConstraint and KernelScale values. The best reported optimization accuracy was 97.73%. A MATLAB GUIDE graphical interface was developed for image upload, feature extraction, prediction, and display of feature-level information. The results indicate that the proposed handcrafted-feature SVM framework is a promising computer-aided screening approach for the analyzed dataset. However, the reported 97.73% value should be interpreted according to the exact validation procedure used in the final code and should not be considered a clinical diagnostic accuracy.