This study delves into “surgomics,” a method that customizes surgical outcome predictions using machine learning and real-time surgical data. The challenge? Gathering high-quality annotations from experts. To address this, the researchers investigated active learning (AL) to make annotation more efficient. They chose ten video-based features linked to intraoperative bleeding complications during robot-assisted esophagectomies. AL significantly improved the recognition of surgical instruments compared to traditional equidistant sampling. This work showcased the potential of AL to streamline the annotation process without compromising machine learning performance in recognizing crucial surgical features.
Journal Article by Brandenburg JM, Jenke AC (…) Wagner M et 22 al. in Surg Endosc
© 2023. The Author(s).