Automated restricted Boltzmann machine classifier for early diagnosis of Parkinson’s disease using digitized spiral drawings

• Published in: Journal of ambient intelligence and humanized computing 2023, Vol.14 (1), p.175-189
• Main Authors: Thakur, Mahima, Dhanalakshmi, Samiappan, Kuresan, Harisudha, Senthil, Ramalingam, Narayanamoorthi, R., Lai, Khin Wee
• Format: Article
• Language: English
• Subjects: Artificial Intelligence; Automation; Biomarkers; Classifiers; Computational Intelligence; Datasets; Diagnosis; Digitization; Disease; Engineering; Handwriting; Kinematics; Machine learning; Motor ability; Multilayer perceptrons; Multilayers; Original Research; Parkinson's disease; Patients; Physical examinations; Robotics and Automation; Signs and symptoms; Skills; Support vector machines; User Interfaces and Human Computer Interaction
• ISSN: 1868-5137; 1868-5145
• DOI: 10.1007/s12652-022-04361-3

Parkinson's disease (PD) is a neurodegenerative disorder that affects the elderly. PD affects the quality of life by causing motor and non-motor disabilities. Traditional PD diagnosis depends on the medical history, a review of symptoms, neurological and physical examinations by a medical specialist. Early detection of PD is a critical step towards providing prompt medical action. In artificial intelligence, computer-assisted methods for PD identification have recently received more attention. The present work focuses on the early detection of PD by logically analyzing time-series data collected during a spiral drawing assessment test of Parkinson’s and normal subjects using digital tablets. A preliminary machine learning approach is taken on static and dynamic drawings tests separately using logistic regression and Support Vector Machine classifier to observe accuracies. It is leveraging a recent novel strategy of employing Restricted Boltzmann machine (RBM) pipelined with multi-layer perceptron model classifier, which provides an accuracy of 95.32% by combining both static and dynamic spiral drawings assessments. The proposed approach is a successful candidate for detecting PD patients. The reported results of cost-effective computer tool-based PD symptom monitoring are helpful in telemedicine applications.