Scalable and Affordable Artificial Intelligence-driven Clinical Decision Support for Movement Disorders - PROJECT SUMMARY/ABSTRACT Parkinson's disease (PD) is the second-most common neurodegenerative condition, affecting over 10 million people worldwide. Misdiagnosis rates of PD range from 15% to 24%. Limited access to PD care is common, especially in rural areas. In the United States (US), only 9% of PD patients see specialists, while most receive care from general neurologists or primary care providers. Mental health care is crucial for PD patients due to the significant burden of psychiatric symptoms, which often go unrecognized and untreated. Barriers include stigma, lack of social support, inadequate screening, and access issues. Artificial Intelligence (AI) shows promise for enhancing neurologists' workflows, predicting outcomes, and developing personalized treatments, thereby improving patient care and quality of life. While multiple studies have quantified motor symptoms in PD using 3D kinematics, video, audio, and wearables, AI integration in clinical practice remains slow due to issues like the black box nature, biases, ethical concerns, and lack of real-world validation. Additionally, research on AI's use in low-resource settings is limited, and there has been little work on quantifying the mental health of individuals with PD in telehealth platforms, which significantly impacts patients’ quality of life. Most studies also suffer from small sample sizes conducted in a research setting. Our proposed research is naturally “translational” as our site records 3D kinematics, multi-view video, and six IMUs along with MDS-UPDRS-III assessment as part of the existing clinical care service (Aim 1). For our long-term vision of AI deployment in academic movement disorder centers, marker-based kinematics measurement during clinical services is billed under Current Procedural Terminology (CPT®) codes for “Medicare” and other insurers. This proposal aims to develop innovative AI for objective assessments of motor symptoms in PD using 3D kinematics, video, and wearable sensors in clinics (Aim 1). We also pioneer accessible multimodal AI for mental health in PD using telehealth platforms (Aim 2) and clinician-in-the-loop AI integration in clinical care settings in an academic movement disorder center (Aim 3), which are important milestones in “clinical AI” in neurology. Our dataset includes a total of 3,085 subjects (1,898 retrospective and 1,187 prospective), arguably the world’s largest database of full-body 3D kinematic behavioral testing data for movement disorders patients with expert review. The outcome of the project has the potential to tackle significant access challenges in monitoring PD in underserved communities through the telehealth system. Our future work will collaborate with the International Parkinson and Movement Disorder Society Telemedicine Study Group to study PD populations in rural U.S. and Global South regions. In partnership with the Jean and Paul Amos Program, we will monitor motor and mental health symptoms and develop evidence-based therapies through telemedicine, interdisciplinary care, and home monitoring. The scalable tools to continuously monitor PD are crucial for timely care to reduce morbidity, disability, and mortality.
