Using artificial intelligence (AI)-based computer vision methods to investigate associations between hand movement and Alzheimer’s Disease (AD) risk

Hand movement can be analyzed by calculating and comparing different movement features such as speed, rhythm, and amplitude. Emerging evidence suggests that hand movement analysis can be used to assess Alzheimer’s disease (AD) risk - a degenerative brain disorder that causes problems with memory, thinking, and behavior, at an early stage, but has never been deeply explored. Extracting hand movement features from videos is a more promising way than from wearable sensors because recording videos are more easily done, and is more scalable to a population level.
Artificial Intelligence (AI)-based computer vision method can be used to analyze hand movement from videos. Markerless hand key point detection is a critical process. By detecting hand key points from the video, the movement of hand key points over a period of time can be used to further extract movement features. However, a hand can move fast and cause motion blur on frames of the video, which may affect hand key point detection performance. Current hand key point detection methods have not been validated in extracting hand movement features from people with AD risk.
First, the thesis reviews the current capabilities of how AI techniques can help assess AD risk. Second, whether current AI-based computer vision approaches can extract accurate hand movement features from videos for further analysis has been investigated. Third, based on the investigation, two novel AI approaches have been proposed to improve hand movement feature extraction from videos. Fourth, an end-to-end hand movement feature extraction system is developed and validated by embedding the aforementioned AI approaches. Fifth, a population-level and online video-based hand movement test is developed and the aforementioned system is applied to extract participant’s hand movement features. Last, a correlation study is conducted to see whether human hand movement is correlated with AD risk.