Diagnosis Parkinson’s Disease Using Neural Network and EEG Signals

Parkinson’s disease (PD) is one of the most prevalent illnesses in existence. There is an increase in the number of illnesses, and the symptoms impair daily activities immediately. By employing EEG signals, our research aids medical practitioners in diagnosing Parkinson’s disease (PD). Our study used a public dataset named EEG: Simon’s Conflict in Parkinson’s on the Open Neuro database. The data includes 28 healthy people and 28 patients with PD (at both ON and OFF medication). Maximal overlap Discrete Wavelet Transform is used to decompose EEG signals into five primary rhythms of EEG and estimate the features. After that, five statistical features, such as mean, kurtosis, skewness, activity, and mobility, are calculated. Several Neural Network models are employed to identify PD. Firstly, we present the classification results of two labels (PD and healthy people) with 88.6% accuracy, 92.0% sensitivity, and 81.7% specificity. Secondly, the performance of classifying three labels (healthy people, PD with ON medication, and of medication) is 83.5%, 74.0%, and 70.5%, respectively. In addition, our research also compared with state-of-the-art papers, and our proposed method achieves outstanding performance.