Future Needle Position Estimation of Suturing Operation using Deep Learning

Abstract

In laparoscopic surgery using a surgical robot, automation of surgical tasks has the potential to increase the efficiency of the surgery including shorten the operation time and reduce fatigue of surgeon. Estimating the position of the suture needle contributes to the semi-autonomous control of the master-follower surgical robot. In this study, we propose a method to estimate the future exit point of the needle from the tissue when suturing with the needle posture before insertion, based on deep learning using the needle segmented. The future tip position of the needle in the 2D image was estimated by neural networks, using past positions obtained from semantic segmentation. The estimation accuracy was compared with neural networks with different structures. We confirmed that the exit point of the needle during suturing can be estimated online within an error of 3mm.