Early detection of Parkinson's disease using data mining techniques from multimodal clinical data

Abstract Parkinson's disease (PD) is the second most common age-related neurological disorder, after Alzheimer's disease. A nervous system specialist, with expertise in sensory system conditions, analyzes the onset of PD, depending on the patient's therapeutic history, an audit of their signs and manifestations, and a neurological and physical examination. Disease diagnosis is done usually based on either voice dataset or spiral drawings. The proposed framework comprises a multimodular arrangement consolidating the static and dynamic spiral tests, and voice datasets for the early discovery of PD. Notwithstanding vocal informational indexes, spiral test illustrations from UCI AI archive are utilized to develop the proposed framework. Individuals experience different side effects at the furthest point of PD. Such being the case, we propose a multimodal approach that improves the reliability of the characteristics of PD-tolerant people. The voice and spiral imaging dataset is trained with standard classifier models. Finally an ensemble-based method is used for disease classification. The final diagnostic decision will be based on either the vocal or spiral ensemble results for detecting PD. Information obtained from investigations into voice and spiral tests is actualized with different AI models like the decision tree, support vector machine (SVM), and Naive Bayes. Performance of the various classifier models was evaluated with the standard accuracy metrics: accuracy, sensitivity, and specificity. The results show that the SVM produces the most accuracy on the spiral imaging datasets, and the proposed ensemble method does the same for disease classification. Early diagnosis of PD helps in immediate treatment planning for people with suspected PD.