AI-DRIVEN INSIGHTS: PREDICTING COGNITIVE DECLINE AND ENHANCING MOTOR FUNCTION RECOVERY IN PARKINSON’S DISEASE THROUGH NEUROIMAGING, BEHAVIORAL DATA, AND ROBOTICS

Abstract

Parkinson's disease (PD) is a neurodegenerative disorder characterized by progressive degeneration of dopaminergic neurons and clinical features of tremor, rigidity, postural instability, and bradykinesia. Early accurate diagnosis remains challenging due to interindividual variability and the inherent limitations of standard imaging modalities. We describe a new multimodal approach integrating neuroimaging, behavioral data analysis, robotics and artificial intelligence. MRI scans were pre-processed and processed using deep learning models—InceptionV3, VGG19, and GANs—to obtain features. The behavioral data was represented using recurrent neural networks (RNNs), while the real-time monitoring of the patient was guaranteed using a robotic-assisted system. The dataset employed is the Parkinson's Progression Markers Initiative (PPMI) and UPDRS-based behavioral recordings. Accuracy, precision, recall, and F1-score metrics were used to compare performance. Overall, AI model achieved 99.5% accuracy in PD classification. CNN-based neuroimaging and RNN-based behavioral data integration improved motor and cognitive decline predictions. Explainable AI revealed salient neuroanatomical features accountable for the model's choices. Robustness was guaranteed by internal cross-validation as well as external validation using independent datasets. This multimodal AI-based approach enhances diagnostic accuracy significantly and facilitates personalized rehabilitation therapy for PD. It has a high potential to bridge the gap between clinical trials and real-world application, paving the way for future developments in neurodegenerative disease management.