Diagnosis of Parkinson's Disease Using Principle Component Analysis and Deep Learning

Parkinson's disease is a common disease which affects the people during the age group of 50's and above. The disease of Parkinson's is the result of damage to the cells which produce a chemical named dopamine in the substantia nigara of the human brain, thus resulting in the failure of the motor movements of the human body, primarily the voice. In this paper, deep learning approach is used to classify healthy people and the people suffering from Parkinson's disease. In the initial step, Principal Component Analysis is used for selecting some suitable features from all the extracted features, thereby reducing the 23 attributes to a mere 11 attributes. The classification of Parkinson's disease is performed with the help of deep learning approach as Artificial Neural Network does not show consistent behavior due to the complex dimensionality of speech data set. The deep learning approach has the capability of to recognize the disease by comparing the various voice signals of the healthy and PD patients. The experiments are performed using 195 voice samples of healthy and Parkinson's disease patients which are retrieved from UCI machine learning repository. The highest accuracy obtained by the proposed approach is 89.23% and sensitivity, specificity of 98.63 and 72.91 respectively. The performance of the proposed approach is compared with existing technique and it is shown that proposed approach is more accurate.