Project Grant 2205454

This $349,683 Project Grant from the National Science Foundation's Computer and Information Science and Engineering program will fund the development of an AI coaching system to enhance surgical teamwork in cardiac operating rooms. Led by William Marsh Rice University, the grant supports the design of multimodal sensing hardware, data-driven algorithms, and an interface to model and computationally generate interpretable feedback and interventions improving teamwork. Researchers will develop a novel Team Markov Model reflecting surgical teams' mental models. Machine learning algorithms using multi-agent imitation learning will arrive at predictive models explicitly dependent on latent performance factors like mental models. Explainable AI techniques will computationally generate interpretable feedback and interventions. Due to challenges collecting large teamwork data sets, the project will emphasize sample- and label-efficient techniques. The integrated system will be prototyped and usability tested using iterative, user-centered design with multi-modal expert-annotated teamwork data in an operating room simulation facility.