Improving the Accuracy of Early Detection of Parkinson's Disease Using Brain Signals Based on Feature Selection in Machine Learning

The diagnosis of Parkinson's disease (PD) is usually done clinically by a doctor. This diagnosis is based on the initial symptoms, motor symptoms, and meditation of the doctor's experience. Since the diagnosis is made with the help of a doctor and based on the clinical description and received information, there is always an error in the diagnosis. Also, early clinical diagnosis is very difficult and almost impossible. Using methods based on machine learning is very useful for early diagnosis of Parkinson's disease. Brain signals and brain function can be a suitable solution for early diagnosis of this disease. Conventional methods are not effective due to the dynamics and complexity of the brain signal. Machine learning methods are a suitable solution with their high capabilities in the process of disease diagnosis. In this article, an efficient method based on machine learning is presented. In this method, after brain signals are pre-processed, time and frequency domain features are extracted from each signal and the best features are selected with the help of the improved intelligent gray wolf algorithm. The selected features are classified using a support vector machine classifier, K nearest neighbor, and random forest. Accuracy higher than 97% shows the superiority of the method in predicting Parkinson's disease.