การพัฒนาแอปพลิเคชันเพื่อคัดกรองผู้ป่วยโรคพาร์กินสัน

Abstract

Parkinson's disease is a complex and progressive neurological disorder that affects patients' motor functions, such as tremors, muscle rigidity, and bradykinesia [1], making daily activities difficult. Diagnosis typically requires a specialist and can be time-consuming. Therefore, the development of a fast and accurate screening tool is essential. This study analyzes patients' handwriting characteristics through tasks such as vertical and horizontal straight-line drawing, writing individual characters, composing Thai sentences, and drawing Archimedean spirals, which reflect abnormalities in motor control. Data collected include pen position (x, y), pressure, and writing duration. These were analyzed in three aspects: bradykinesia, muscle rigidity, and micrographia [2]. Since the extracted features from position, pressure, and duration across different tests can be extensive, statistical analysis was applied to select appropriate features using the ReliefF algorithm combined with Sequential Forward Selection (SFS). The selected features were then classified using models such as Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Decision Tree. Model performance was evaluated using repeated 10-fold cross-validation for 50 iterations [3] [4]. Results showed that using SFS with SVM yielded the highest accuracy of 84.91%, with a sensitivity of 78.05% and specificity of 90.96%. This indicates that handwriting analysis can serve as an effective tool for Parkinson’s disease screening, offering faster and more accurate detection, reducing diagnostic delays, and enabling timely treatment for patients.