The Trauma THOMPSON Dataset for Real-World Emergency AI


Yupeng Zhuo (Purdue University), Eddie Zhang (Purdue University), Xiangchen Yu (Purdue University), Aditya Pachpande (Purdue University), Andrew Wallace Kirkpatrick (University of Calgary), Kyle Couperus (The Geneva Foundation), Jessica Mckee (University of Calgary), Juan Wachs (Purdue University)
The 35th British Machine Vision Conference
PDFPosterVideo (Right click to download)

Abstract

We introduce the Trauma THOMPSON dataset, a large-scale resource aimed at advancing artificial intelligence for real-time decision support in emergency and austere medical environments. The dataset consists of 220 unscripted, egocentric recordings of five emergency procedures, including a diverse set of “just-in-time” (JIT) life-saving interventions carried out under resource-limited conditions. These scenarios capture the realities of humanitarian and field medicine, where practitioners must frequently adapt or improvise beyond standard protocols. To enable richer visual understanding, we provide two new layers of fine-grained annotations: bounding boxes for critical medical instruments and supplies, and detailed hand annotations to support tracking and surgical skill assessment. Together, these enable research on spatiotemporal reasoning, interaction modeling, and the development of AI copilots capable of interpreting and assisting with complex procedures in real time. The dataset supports multiple benchmark tasks, including action recognition, action anticipation, visual question answering (VQA), and object detection. We benchmark state-of-the-art models across these tasks, highlighting both current strengths and the challenges of building reliable AI systems for deployment in the field. The dataset is publicly available at https://doi.org/10.7910/DVN/V5BTRU and provides a foundation for the creation of intelligent systems that can assist frontline medical personnel.

Citation

@inproceedings{Zhuo_2025_BMVC,
author    = {Yupeng Zhuo and Eddie Zhang and Xiangchen Yu and Aditya Pachpande and Andrew Wallace Kirkpatrick and Kyle Couperus and Jessica Mckee and Juan Wachs},
title     = {The Trauma THOMPSON Dataset for Real-World Emergency AI},
booktitle = {36th British Machine Vision Conference 2025, {BMVC} 2025, Sheffield, UK, November 24-27, 2025},
publisher = {BMVA},
year      = {2025},
url       = {https://bmva-archive.org.uk/bmvc/2025/assets/papers/Paper_1135/paper.pdf}
}


Copyright © 2025 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection