A Hybrid Transfer Learning Approach to Detect Parkinson's Disease

Parkinson's disease (PD) is a degenerative brain condition characterized by motor and non-motor syndromes. Patients often experience hand tremors, affecting writing and drawing skills and slow motor action. Early detection of PD can improve medical treatment and lifestyles. Hence, this study aims to use a hybrid approach comprised of VGG16 Convolutional Neural Network and Random Forest (RF) algorithm to detect PD from archimedean spiral and wave drawing image datasets. The number of training datasets has been increased by utilizing image augmentation to train the model accurately. The accuracy of the model for the spiral drawing dataset is 93.34%, and for the wave drawing dataset, it is 90%.