AI System Aims to Improve Early Parkinson’s Detection

New technology uses smartphone videos to analyse movement

A research project at the University of Bradford is developing an artificial intelligence (AI) system designed to assist in the early detection and monitoring of Parkinson’s disease by analysing movement through smartphone video recordings. The initiative involves collaboration with clinicians from Leeds Teaching Hospitals NHS Trust and Hospital de Clínicas in Paraná, Brazil.

Understanding Parkinson’s Disease

Parkinson’s is a progressive neurological disorder characterised by the gradual loss of dopamine-producing neurons in the brain. This leads to symptoms affecting movement and coordination, such as tremors, muscle rigidity, impaired balance, and bradykinesia, which is a slowness of movement. As the condition advances, individuals may lose functional independence, impacting their ability to perform daily tasks.

Currently, Parkinson’s assessments are conducted by neurologists who use the Movement Disorder Society revision of the Unified Parkinson’s Disease Rating Scale (MDS-UPDRS). This involves a 50-question evaluation covering motor and non-motor symptoms, with clinicians relying on visual interpretation of motor signs like bradykinesia. These assessments require experienced clinicians and sustained patient engagement.

How the AI System Works

The AI system developed by the Bradford team uses computer vision technology to analyse videos of patients performing specific movement tasks. These tasks include finger tapping, wrist rotation, hand opening and closing, and heel or toe tapping while seated. Videos are recorded on smartphones and uploaded to a cloud platform for analysis.

• The system applies AI models, including Google’s MediaPipe and custom-trained algorithms, to extract detailed motion metrics.
• It classifies movement severity on the same five-point scale used by neurologists in clinical assessments.
• By converting subjective clinical judgements into objective data, the system aims to track changes in Parkinson’s symptoms over time.

Assessing Fall Risk Through Movement

Since Parkinson’s is associated with an increased risk of falls, the research also focuses on evaluating lower-limb function, particularly ankle performance. One test involves a 30-second ‘up on the toes’ movement, which can be safely performed in a small space without specialist supervision.

The AI analyses consistency and performance in this test, potentially offering an alternative to traditional gait analysis methods that require more space, equipment, and clinical oversight. Videos for this assessment are also captured via smartphone and processed remotely.

Research Progress and Future Directions

Since 2019, trials involving 120 participants have utilised finger pinching and up-on-the-toes tests, showing potential for both Parkinson’s detection and fall risk identification in older adults. The system is intended not only for early screening but also for ongoing monitoring, allowing clinicians to track symptom progression or treatment response through regular home recordings.

While the Leeds section of the research has paused, collaborations in Brazil continue, with new patient data contributing to system development. The research team hopes that with further support, the technology could advance towards practical deployment, enabling faster diagnosis, consistent monitoring, and earlier intervention for Parkinson’s disease.