Enhanced Machine Learning Models for Parkinson’s Disease Detection Using Spiral Drawing Data: A Comprehensive Study

Abstract Background Parkinson's disease (PD) is a progressive neurodegenerative disorder that severely affects patients' quality of life. Early and accurate diagnosis is essential for timely intervention. Traditional diagnostic approaches rely heavily on subjective clinical evaluation. Objective This study aims to evaluate various machine learning (ML) models for PD detection using spiral drawing data, focusing on preprocessing, feature extraction, and statistical validation. Methods The dataset, sourced from the UCI Machine Learning Repository, includes 122 samples (61 PD patients and 61 controls). Five ML classifiers were evaluated: Support Vector Machine (SVM), Random Forest (RF), K-Nearest Neighbors (KNN), Logistic Regression (LR), and Gradient Boosting (XGBoost). Preprocessing included normalization, Principal Component Analysis (PCA), and SelectKBest. Model evaluation was done using 10-fold cross-validation, hyperparameter tuning with GridSearchCV, and metrics such as accuracy, precision, recall, F1-score, ROC, and AUC. Results Gradient Boosting achieved the highest performance (91.2% accuracy, AUC 0.95). Dimensionality reduction and feature selection significantly enhanced model performance, particularly for SVM and KNN. Conclusion Advanced ML techniques, when combined with proper preprocessing, show significant promise in PD diagnosis using spiral drawing data. Future research should explore deep learning and multimodal integration.