AI-based video diagnostic software for Parkinson's Disease.


Over 10 million people in the world are living with Parkinson’s disease (PD), a disease which causes over 29,000 deaths per year and affects around 900,000 people in the United States alone. Shockingly, even in developed nations like the United States, an estimated 40% of people aged 65 or older living with PD do not receive care from a neurologist. A large problem in both Western and Asian countries is that many individuals, around 15% or higher, living with PD have not been diagnosed. Even when diagnosed, PD requires a continuous process of trying and adjusting medications to reduce symptoms to improve their quality of life. This involves frequent appointments with a neurologist, a costly and time-intensive process.



We have developed an online tool, called PARK (, for analyzing the symptoms of Parkinson’s disease from video and audio information. PARK instructs and guides users through motor and audio tasks selected from the standardized Movement Disorder Society Unified Parkinson's Disease Rating Scale (MDS-UPDRS). While performing the tasks, PARK records the audio and video via webcam and microphone. PARK allows users to navigate the system using only the space-bar from the keyboard, making it accessible to users with movement disorders. The recordings are uploaded to a secure server where the motion, facial, and audio features are extracted and analyzed using machine learning models. These models have been trained using extracted features from 600 users (400 PD, 200 non-PD) and are able to predict whether a person is likely to have Parkinson’s disease.


In addition to recording and processing, PARK allows individuals to review the analysis. PARK shows the tremor level of the user’s hand in a bar plot, involuntary head movement frequency in a line graph, and delay in uttering the same sentences in numeric values. In addition to the visuals, PARK gives a suggested probability of having PD on a scale of 0 to 100, allowing users to determine if they should set an appointment with a neurologist to obtain a definitive diagnosis. Users can log in to PARK using their email to track their disease/symptoms progression over time and have the option to share the analysis results with their neurologists/physicians through email.

URV Reference Number: 1-21017
Patent Information:
For Information, Contact:
Curtis Broadbent
Licensing Manager
University of Rochester
Mohammed Ehsan Hoque
Earl Ray Dorsey