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

Abstract

Parkinson's Disease (PD) is difficult to identify early on since there are no physical diagnostics. This initiative addresses the critical need for a non-invasive Parkinson's disease detection method. This study uses deep learning, primarily CNNs, to discover Parkinson's disease patients by analyzing handwriting trends. In classification jobs, deep learning has shown to be very accurate, and it has also been useful in medical areas for analyzing data like X-rays and MRI scans. The project's main goal is to improve accuracy by taking traits from handwriting, teaching machine learning models, and checking how well they work against old-fashioned methods. This project boosts its skills by adding the Xception algorithm for better feature extraction, in addition to VGG19, InceptionV3, and ResNet50 for feature extraction. A strong voting algorithm is added to the classification process to make it stronger. A Flask system that works with SQLite has been created to make user testing and real-world use easier. This project not only makes the project bigger, but it also makes sure that the interface is easy to use so that it can be tested and proven in clinical situations.