Objective: We aimed to demonstrate that an AI-powered virtual exam using a standard webcam can assess PD manifestations, naturally mirroring a neurologist’s evaluation.
Background: Parkinson’s disease (PD) motor evaluations remain primarily clinical, making them subjective and prone to recall bias. Developing objective tools is essential for accurate diagnosis and disease progression tracking.
Method: We studied people with PD (PwP) diagnosed according to the MDS criteria and healthy controls (HC). AI-driven modules processed webcam signals to automate a virtual neurological exam, specifically a structured neurological examination performed in an uncontrolled environment. Tasks included upper limb maneuvers (finger tapping, hand grasp, and pronation-supination), lower limb assessments (heel/toe taps), and gait analysis over a 10-meter corridor. Metrics included amplitude, frequency, fatigability, speed, and movement interruptions. Concurrent validity was assessed against traditional clinical scales.
Results: We evaluated 28 PwP (median age 68.2, UPDRS III: 37, Levodopa equivalent dose 443mg) and 9 HC (median age 64.2, UPDRS III: 0). Key discriminators included: amplitude of toe tapping (p=0.006) and finger tapping (p=0.004); coefficient of variation for toe (p=0.004), finger (p=0.003), and heel tapping (p=0.004); duration of heel (p=0.016) and finger tapping (p=0.030) and their respective coefficients of variation (p=0.036, p=0.016); and heel tapping interruptions (p=0.004). Regarding gait, PwP had a longer step duration (+700 ms), reduced stride length (−10 cm left, −8 cm right), and lower velocity (−2 cm/s, −4 cm/s) than HC. Various AI-generated metrics moderately correlated (Spearman Rho >0.5) with MDS-UPDRS III scores.
Conclusion: AI-driven webcam assessments seamlessly and effectively differentiate PwP from HC when applied to unobtrusively recorded traditional evaluations. Further studies with larger cohorts and refined AI models could establish this as a robust, objective PD assessment tool that could support current clinical practice.
To cite this abstract in AMA style:
J. Lapeña-Motilva, K. Coutinho-García, E. Rangel, JC. Martínez-ávila, D. Pérez-Martinez, M. Hernández Gonzalez-Monje, N. Malpica-González, A. Sánchez-Ferro. AI-Powered Seamless Webcam Assessment for Parkinson’s Disease [abstract]. Mov Disord. 2025; 40 (suppl 1). https://www.mdsabstracts.org/abstract/ai-powered-seamless-webcam-assessment-for-parkinsons-disease/. Accessed October 5, 2025.« Back to 2025 International Congress
MDS Abstracts - https://www.mdsabstracts.org/abstract/ai-powered-seamless-webcam-assessment-for-parkinsons-disease/