This National Science Foundation (NSF) Computer and Information Science and Engineering (CISE) Federal Grant Award (CFDA 47.070) provides $1,183,690 to The Regents of the University of California, San Francisco (UCSF) to develop personalized machine learning models that can predict repeat adverse health events like substance use and stress-related hypertension using data from consumer wearable devices.
The project aims to create artificial intelligence (AI) models that can learn from an individual's passively collected wearable sensor data, such as heart rate and movement patterns, to predict the occurrence of complex health outcomes like substance use and blood pressure spikes. The innovation lies in leveraging self-supervised learning techniques to train these personalized AI models on large amounts of unlabeled user data, requiring fewer labeled examples to make accurate predictions compared to traditional approaches. The project will test this paradigm through two user studies targeting stress-related hypertension and substance use detection. Additionally, the researchers plan to develop human-computer interaction techniques to facilitate efficient annotation of adverse health events by end-users, further enhancing the practical utility of the personalized AI models. This 4-year project, which commenced on January 1, 2025, represents NSF's continued investment in advancing AI and machine learning capabilities with the potential for significant health impact.
Generated 3/25/25, 3:27 AM