A Novel Approach on Parkinson Detection with Deep Ensemble Network
Anantha Sivaprakasam S, Senthil Pandi S, Varsha SP, M Vishnu, Japa Sai Sharath
Abstract
Parkinson Disease (PD) is one of the serious and life- threatening disease treated by doctors with various progressive diagnosis. Magnetic resonance imaging (MRI) plays a significant role in treating the disease in the early stages. Although various screening techniques are utilized for PD, MRI technique provides accurate extraction of Parkinson features impacting the disease. The emerging growth of artificial intelligence frameworks in recent days create a strong foundation in medical image processing and critical disease inference. The presented system formulated with such framework using an ensemble approach. The MARI images are collected from UCI repository. The involvement of image processing technology and deep learning, deep ensemble network (DEN) is developed here. The primary goal of the system is to detect the presence of Parkinson disease in the early stages. The proposed approach is created with deep ensemble process with Convolutional Neural Network (CNN) and Long-Short Term Memory model (LSTM) with ADAM optimizer. The performance of the system is evaluated using accuracy, precision, recall and F1score. Through DEN model, accuracy of 98% is achieved. The quantitative parameters are compared with the existing state of art approaches.