Enhancing Parkinson’s Disease Diagnosis Through Image-Driven Deep Transfer Learning and Optimization

Abstract

This research proposed a hybrid methodology that combines data augmentation approaches, feature extraction using a pretrained Convolutional Neural Network (CNN), feature selection through optimization, and classification through Machine Learning to enhance the identification of Parkinson's Disease. This research first uses six different pretrained CNN models to classify different types of handwriting images (circle, spiral, and meander). The VGG16 framework works better than the others. The second step uses Binary Grey Wolf Optimization (BGWO) to choose the best collection of features from the VGG16 network by freezing the layers. The suggested strategy gets a 99.8\% accuracy rate in classification using Support Vector Machine (SVM). We used the NewHandPD benchmark dataset to test how well our technique works. The experimental results demonstrate that the proposed method diagnoses Parkinson's disease more effectively.