Lapex: A new multimodal dataset for context recognition and practice assessment in laparoscopic surgery

The LapEx dataset contains 30 videos of sleeve gastrectomy surgeries performed by two surgeons. The videos were captured at 25 fps. Three annotation tasks were performed on the videos of the "dissection of fundus" surgical step : 1. Surgical activites were annotated with the quadruplet < actor , instrument , verb , target > 2. The quality of exposure was annotated with a value in [good, satisfying, unsatisfying] 3. Images were completely segmented with a labeling of visible tools and organs

LapEx数据集包含由两名外科医生完成的30例袖状胃切除术手术视频，所有视频均以25帧/秒（fps）的帧率录制。针对“胃底分离”这一手术步骤的视频，共完成了三项标注任务：
1. 手术活动以<参与者（actor）、手术器械（instrument）、操作动作（verb）、靶标组织（target）>四元组形式进行标注；
2. 手术视野暴露质量以[良好（good）、满意（satisfying）、不佳（unsatisfying）]三个等级进行标注；
3. 对所有视频帧图像进行全分割，并标注其中可见的手术器械与器官。

搜集汇总

背景与挑战

背景概述

Lapex数据集是一个多模态腹腔镜手术数据集，包含30个袖状胃切除术视频，重点对'胃底解剖'步骤进行了三重注释（手术活动、暴露质量和图像分割），适用于手术场景识别和实践质量评估研究。