AutoLaparo
收藏资源简介:
AutoLaparo是香港中文大学机械与自动化工程学系开发的首个集成多任务数据集,专注于腹腔镜子宫切除手术的图像引导自动化。该数据集包含21个完整的子宫切除手术视频,总时长1388分钟,涵盖手术流程识别、腹腔镜运动预测和器械及关键解剖结构分割三个任务。数据集通过精细的三级标注过程创建,旨在通过多任务和多模态学习提升手术场景的高级感知能力,推动图像引导手术自动化的发展。
AutoLaparo is the first integrated multi-task dataset developed by the Department of Mechanical and Automation Engineering, The Chinese University of Hong Kong, focusing on image-guided automation for laparoscopic hysterectomy surgery. This dataset contains 21 complete hysterectomy surgery videos with a total duration of 1388 minutes, covering three tasks: surgical workflow recognition, laparoscopic instrument motion prediction, and segmentation of surgical instruments and key anatomical structures. Developed via a meticulous three-level annotation process, this dataset aims to enhance advanced perception capabilities in surgical scenarios through multi-task and multi-modal learning, and promote the development of image-guided surgical automation.




