Automatic Surture thread detection for surgical assistance

Abstract

Surgical environment recognition is indispensable for automatic robotic assistance. In particular, the surgical thread’s detection is essential to the automation of micro-anastomosis but challenging because of the small diameter of the thread. This work investigates the use of an image-based structural health monitoring deep learning algorithm and motion blur augmentation to improve surgical thread recognition. The method splits the image into a grid of sub-images and judges whether each sub-image is part of the surgical thread. With that information, it reconstructs the location of the thread in the image. The proposed method identified the surgical thread area with a degree of success but misidentified a part of the boundary.