V-JEPA, HeiCo dataset, in-house dataset
收藏资源简介:
本研究使用了三个数据集:V-JEPA,HeiCo数据集和内部数据集。V-JEPA是一个在自然视频场景上训练的多模态模型,用于微创手术支持。HeiCo数据集提供了三种类型手术的腹腔镜视频和手术室医疗设备状态数据,以及每个视频帧的14个手术阶段标签。内部数据集包含了来自不同患者的腹腔镜视频和手术过程中的4个生命体征数据流,以及患者术后住院天数和并发症影响标签。这些数据集用于分析模型在预测医院住院时长和术后并发症方面的表现。
This study utilizes three datasets: V-JEPA, the HeiCo Dataset, and an internal dataset. V-JEPA is a multimodal model trained on natural video scenarios for minimally invasive surgery support. The HeiCo Dataset provides laparoscopic videos of three types of surgeries, operating room medical equipment status data, and 14 surgical phase labels for each video frame. The internal dataset contains laparoscopic videos from different patients, four vital sign data streams collected during surgery, as well as labels for patients' postoperative hospital stays and the impact of complications. These datasets are used to analyze the performance of models in predicting hospital length of stay and postoperative complications.
数据集概述
基本信息
- 标题:Leveraging generic foundation models for multimodal surgical data analysis
- 作者:Simon Pezold, Jérôme A. Kurylec, Jan S. Liechti, Beat P. Müller, Joël L. Lavanchy
- 年份:2025
- DOI:10.48550/arXiv.2509.06831
内容描述
- 该数据集为论文《Leveraging generic foundation models for multimodal surgical data analysis》的代码仓库。
- 包含代码、模型权重和使用说明,即将发布。
引用信息
bibtex @article{pezold2025leveraging, title = {Leveraging Generic Foundation Models for Multimodal Surgical Data Analysis}, author = {Pezold, Simon and Kurylec, Jérôme A. and Liechti, Jan S. and Müller, Beat P. and Lavanchy, Joël L.}, journal = {arXiv preprint}, year = {2025}, doi = {10.48550/arXiv.2509.06831}, }




