Hybrid of Deep Learning and Transformer Models with Explainable IA for Enhanced Parkinson's Disease Detection

Parkinson's disease is an incurable neurodegenerative disease mostly common to the senior citizen. Currently it affects nearly one million people in United state and over 10 million people worldwide. Traditionally, Parkinson's disease has been diagnosed by physicians through medical examinations. However, with advancements in technology, we have explored the use of hybrid deep learning and Transformer models for the prediction and detection of Parkinson's disease. We developed six models using a well-balanced dataset consisting of approximately 1,040 instances, with 520 instances each from PD patients and healthy controls. LIME was applied to each prediction to provide explanations for the model's outputs. Remarkably,. four out of the six hybrid models achieved 100% accuracy, delivering trustworthy and reliable predictions.