Instance Segmentation of Laparoscopic Instruments and Organs

Abstract

Robot assisted laparoscopic surgeries gained more light in the past few years especially after Da Vinci has been introduced in the medical field. Advancement in AI techniques and robotic procedures led scientist to the advancement in medical surgeries by automate the surgical procedures. In the presented work, we propose a novel segmentation algorithm for identification, label and classification of the tissues, organs and surgical tools in the endoscopic video feed of the human torso region. This thesis serves as the first step towards autonomous minimal invasive surgery. It has two main contributions: first, we contribute an annotated dataset called M2CAISeg created from actual endoscopic video feed of surgical procedures, and secondly, we propose a state of the art deep learning algorithm for instance segmentation. The trained model will be cross-validated followed by comprehensive evaluation on the test set of the proposed dataset.