An Intelligent System for Parkinson's Diagnosis Using Hybrid Feature Selection Approach

Parkinson's is the second most common neurodegenerative disorder after Alzheimer's disease. During the nascent stage, the symptoms of Parkinson's disease are mild and sometimes go unnoticed, but as the disease progresses, the symptoms become severe. Recent research has shown that changes in speech or distortion in voice can be effectively used for early Parkinson's detection. In this work, the authors propose a system of Parkinson's disease detection using speech signals. As the feature selection plays an important role during classification, the authors have proposed a hybrid MIRFE feature selection approach. The result of the proposed feature selection approach is compared with the five standard feature selection methods by XGBoost classifier. The proposed MIRFE approach selects 40 features out of 754 features with a feature reduction ratio of 94.69%. An accuracy of 93.88% and area under curve (AUC) of 0.978 is obtained by the proposed system.