Multivariate vector autoregressive prognosis-based model following control method for robot-assisted beating heart surgery

Abstract

Robotics tool enables the surgeon to conduct the operation on the beating heart during the off-pump coronary artery graft bypass surgery. The robotic tool actively cancels the relative motion between the point of intersect (POI) on the surface of the heart and surgical tool, which allows the surgeon to operate as if the heart is stationary. The nonlinear nature of the beating heart motion disables the conventional feedback controller and pose difficulty to the robot tool. We apply Granger causality to analyze simultaneously measured electrocardiography (ECG) and 3D heart motion data. The extracted interdependency between the different sets of time series reveals the feasiblity for the improved prediction of the heart motion. In this paper, we propose an adaptive multivariate vector autoregressive (MVAR) prognosis-based model following (MF) control algorithm. Using this method, the prediction part takes advantage of the ECG signal, which contains the nonstationary heart rate dynamics significantly correlated with the heart motion. In addition, the control part automatically incorporates the feedforward path to enhance the tracking of the abrupt change and occasional abnormality in heart motion. The comparative experiment results for evaluating the proposed algorithm are reported by using the vivo collected data. The results indicate that MVAR-based MF improves the control accuracy by 0.2 mm and achieves better tracking performance by capturing more nonlinear characteristics of the heart motion, and following the heart motion better with sufficient details.