Transparent AI Solutions for Parkinson’s Disease Diagnosis and Monitoring: A Multimodal Deep Learning Approach

Parkinson’s disease (PD) diagnosis remains challenging due to the variability in motor and non-motor symptoms. While deep learning (DL) models show promise in improving diagnostic accuracy, their "black box" nature limits clinical adoption. This study proposes an explainable AI (XAI) framework combining DL models with interpretability techniques, such as SHAP and LIME, to enhance transparency and trust. Using multimodal datasets, including clinical, imaging, and wearable sensor data, the framework identifies key features driving predictions, enabling clinicians to make informed decisions. The proposed method achieves superior performance, with an accuracy of 92%, precision of 91.1%, and recall of 90.4%, outperforming existing methods. By bridging the gap between AI outputs and clinical needs, this framework offers a practical tool for early PD diagnosis and progression monitoring.