SCH: BRIDGEX: Bridging the Gaps in Robotic Surgery Knowledge Sharing Through Explainable Augmentation and Learning - PROJECT SUMMARY (See instructions): Robotic-assisted surgery has improved precision, visualization, and minimally invasive care, but the safe translation of artificial intelligence into surgical practice remains limited by gaps in knowledge sharing, interpretability, and data access. Surgeons contribute expertise in anatomy, workflow adaptation, and intraoperative decision-making, while engineers develop robotic control, computer vision, simulation, and Al methods; however, these perspectives are rarely integrated into a shared computational framework. In addition, multi modal surgical data, including endoscopic video, robotic kinematics, tool trajectories, tissue deformation, simulation data, and clinical annotations, remain distributed across institutions because of privacy, governance, infrastructure, and data-standardization constraints. The long-term objective of this project is to establish an explainable and privacy-preserving Al framework that supports collaborative knowledge sharing and model development for robotic surgery, with the goal of advancing safer, more interpretable, and more generalizable Al-assisted surgical technologies. The proposed framework, BRIDGEX, will address this objective through three specific aims. Aim 1 will develop biomechanically grounded vision-language representations that encode surgical video, robotic motion, tissue deformation, and simulation-based constraints in a shared latent space. Aim 2 will develop trustworthy federated learning methods that enable cross-institutional model training and adaptation without direct sharing of raw patient data. Aim 3 will develop an interactive retrieval-augmented language model interface that provides evidence-grounded explanations for surgeons and engineers by linking Al outputs to surgical context, robotic trajectories, and simulation-based reasoning. The research design will integrate algorithm development with simulation studies, phantom experiments, robotic testbeds, and pilot workflow studies involving the University of Tennessee, Knoxville and the University of Tennessee Medical Center. Successful completion is expected to produce new methods for privacy-preserving surgical Al, explainable human-Al interaction, and clinically grounded robotic surgery decision support.
