Neural network approaches reduce RMSE by 60% in robotic-assisted beating heart surgery motion compensationRobotic Surgery Technology Shows Promise for Beating Heart Procedures

Robotic-assisted minimally invasive cardiac surgery offers the potential to perform off-pump beating heart procedures without cardiopulmonary bypass, but active closed-loop algorithmic motion compensation, as distinct from image-stabilization, motion-tracking, and mechanical-stabilization approaches that have seen limited clinical use, has not yet entered human clinical trials despite two decades of research. This scoping review, conducted using the PRISMA-ScR framework, addresses three questions: what are the quantitative performance limits of current sensing, prediction, and control methods; what is the state of clinical translation; and what barriers prevent first-in-human deployment. Studies were included if they reported methods or quantitative results for sensing, predicting, or compensating cardiac motion in robotic surgery (61 studies included); non-robotic stabilization studies and non-English publications were excluded. Four databases were searched from 1998 to May 2026; formal risk of bias assessment was not performed, consistent with PRISMA-ScR guidance. Results were synthesized narratively by thematic area. Key results include: dual Fourier-series EKF achieves 0.73–0.81 mm RMSE at 180 ms horizon; neural network approaches reduce this by 60% at 160 ms latency; cascade MPC-AOB maintains