Parkinson Disease Detection using Feed Forward Neural Networks

This paper is based on detection of Parkinson's Disease (PD) a chronic disease which is recognized as the second most progressive neurological disorder. PD is characterized by symptoms including tremors, bradykinesia, rigidity, blepharospasm, dysphonia, postural instability difficulty with balance and coordination and slowness of movement. As the exact cause of Parkinson's disease is not fully understood, but the ongoing research claim that PD patients have dysphonia at early stage i.e., voice impairments. There will be a greater interest in machine learning researchers that detecting voice changes in PD patients which makes the early detection is accomplished by using Machine Learning techniques. This work is continuation to the paper titled” Diagnosis of Parkinson's Disease using Principal Component Analysis and Machine Learning algorithms with Vocal Features” [1]. In this work, we propose Feed Forward Neural Networks Algorithm to increase the Accuracy of detecting Parkinson disease. The dataset used here is the voice recordings of both PD and Healthy Conditioned(HC) patients, that are extracted to speech features and is taken from UCI repository. The Evaluation metrics such as accuracy, Precision, Recall, F1 score, Area Under ROC(Receiver operating Characteristic) curve are used for evaluating the Feed Forward Neural Networks model.