Feasibility of infrared tracking of beating heart motion for robotic assisted beating heart surgery

A comprehensive multimodality heart motion prediction algorithm for robotic-assisted beating heart surgery

BACKGROUND

An essential requirement for performing robotic-assisted surgery on a freely beating heart is a prediction algorithm that can estimate the future heart trajectory.

METHOD

Heart motion, respiratory volume (RV) and electrocardiogram (ECG) signal were measured from two dogs during thoracotomy surgery. A comprehensive multimodality prediction algorithm was developed based on the multivariate autoregressive model to incorporate the heart trajectory and cardiorespiratory data with multiple inherent measurement rates explicitly.

RESULTS

Experimental results indicated strong relationships between the dominant frequencies of heart motion with RV and ECG. The prediction algorithm revealed a high steady state accuracy, with the root mean square (RMS) errors in the range of 82 to 162 μm for a 300-second interval, less than half of that of the best competitor.

CONCLUSION

The proposed multimodality prediction algorithm is promising for practical use in robotic assisted beating heart surgery, considering its capability of providing highly accurate predictions in long horizons.

A Hybrid Algorithm for Prediction of Varying Heart Rate Motion in Computer-Assisted Beating Heart Surgery

An essential requirement for performing robotic assisted surgery on a freely beating heart is a prediction algorithm which can estimate the future trajectory of the heart in the varying heart rate (HR) conditions of real surgery with a high accuracy. In this study, a hybrid amplitude modulation- (AM) and autoregressive- (AR) based algorithm was developed to enable estimating the global and local oscillations of the beating heart, raised from its major and minor physiological activities. The AM model was equipped with an estimator of the heartbeat frequency to compensate for the HR variations. The RMS of the prediction errors of the hybrid algorithm was in the range of 165-361 μm for the varying HR motion, 21% less than that of the single AM model. With the capability of providing highly accurate predictions in a wide range of HR variation, the hybrid model is promising for practical use in robotic assisted beating heart surgery.