Deep learning based Surgical Workflow Recognition from Laparoscopic Videos

Technical evolution within the surgical healthcare environment resulted in transformative and innovative surgical advancements in biomedical science, which require contemplation for further renovation. Systematizing of surgical workflow has immense potential in offering solutions to numerous stipulations of the modern operating room. Cataloging surgical phases are vital in developing context-aware systems that can supervise the surgical procedures, optimize the operating room environment, scheduling of medical staff, and provide automated assistance to the clinical personals. This research has mainly been focused on a deep learning approach to sense and detect phases of a surgical operation. Classification of surgical phases is based on the CNN architecture ResNet50 and performs the rolling average prediction method to accurate the results. The proposed framework automatically recognizes the various phases in the laparoscopic cholecystectomy videos reliably.