Deep Transfer Learning Based Parkinson’s Disease Detection Using Optimized Feature Selection

Abstract

The absence of the conclusive medical diagnostics complicates the prognosis of the Parkinson Disease (PD) and particularly during the initial stages of the disease. The proposed study addresses the urgent desire to have a green and non-invasive approach to early PD identity through the application of “deep learning, in particular, Convolutional Neural Networks (CNNs)” to analyze handwriting patterns. Various fashions are employed to extract features such as “ResNet50, VGG19, Inception V3, and Xception and K-Nearest Neighbours (KNN), Random Forest (RF), Support Vector Machine (SVM), and Decision Tree” are performed to extract type. The warning ensemble strategy combines predictions of numerous fashions thus improving accuracy. The main one, i.e. “ResNet50, VGG19, and InceptionV3 using KNN”, achieved 95% accuracy. An additional study of ensemble techniques that incorporate the “Voting Classifier” is conducted with the aim of hitting 98 per cent and beyond. A front end the usage of the flask framework has been set up to be used by person trying out, including user authentication. The finding supplements the preliminary identification of the sickness of Parkinsonism, which is vital in administering timely treatment and enhancing satisfactory of lifestyles by the sufferers.