Image Segmentation for Robotic Microsurgery

―Evaluation of segmentation methods for a surgical suture―

抄録

Robot-assisted microsurgery is laborious due to the restricted operation space and the tiny size of the surgical objects. To alleviate the workload, it is necessary to improve the autonomy of the surgical robot, in which the robot's objects recognition plays the key role. Recently, Image segmentation has been widely studied in the field of objects recognition because it provides abundant visual information, however, segmentation of small and thin objects in surgical microscopic view such as needle and thread can be really challenging. In this paper, image segmentation methods including the Transformer have been evaluated to detect a surgical suture consisting of needle and thread. The result shows that Transformer can achieve a better segmentation for needle and thread.