Abstract
Bleeding is one of the most frequent and clinically significant events affecting surgical safety. In both intraoperative risk control and postoperative assessment, blood loss is closely associated with patient safety and mortality, surgical quality, and treatment decisions. Traditional methods for blood loss assessment, such as suction canister volume, only provide a global, postoperative estimate and thus fail to capture temporal characteristics and local risks of bleeding. To address these limitations, we propose two video analysis tasks related to surgical bleeding. The first
task is clip-level bleeding severity classification. We constructed a newly annotated dataset containing 66 videos of robotic liver resections under a
slightly modified version of the Validated Intraoperative Bleeding (VIBe) scale. This task aims to classify the bleeding severity of short surgical video
clips, enabling fine-grained analysis of intraoperative bleeding conditions. The second task is surgery-level blood loss estimation for the entire surgery. Inspired by the blood loss component of the Surgical Apgar Score, we divide the total blood loss of a surgery into different severity levels and use clip-level bleeding predictions to estimate the overall blood loss category of the full surgical video. We evaluate several mainstream video recognition models on the proposed tasks and further introduce Bleeding Severity MViT, a MViT-based model designed for bleeding severity classification. Our model achieves the best performance among the evaluated baselines, particularly in recognizing more severe bleeding cases. We also explore a two-stage framework combining a binary bleeding classifier with a bleeding severity classifier, which demonstrates the potential to outperform end-to-end models. Finally, we apply our model to unlabeled surgical videos and validate its effectiveness for surgery-level blood loss estimation with an accuracy of 56.1%.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/SafeSurg_027.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: https://papers.miccai.org/miccai-2026-sat/supp/SafeSurg_027_supp.pdf
Link to Open Review
Open Review Page: https://openreview.net/forum?id=2R4rBJlkp7
BibTex
@InProceedings{LiPen_Hemorrhage_MICCAISAT2026,
author = { Li, Pengfei AND Kurylec, Jérôme A. AND Girish, Rao Rohan AND Kannathasan, Aatharshan AND Müller, Philip C. AND Müller, Beat P. AND Pezold, Simon AND Lavanchy, Joël L.},
title = { { Hemorrhage First: Rethinking Surgical Safety Assessment via Video-based Blood Loss Estimation } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS pending},
month = {pending},
page = {pending}
}