Synergistic Bleeding Region and Point Detection in Laparoscopic Surgical Videos

TL;DR

This work tackles automated detection of intraoperative bleeding by introducing SurgBlood, a real-world dataset with region and point annotations, and BlooDet, a dual-task online detector built atop SAM 2. BlooDet uses a mask branch with an edge generator and a point branch with memory modeling to jointly detect bleeding regions and localize bleeding points, guided by cross-branch prompts and temporal cues from optical flow. The model employs an alternating optimization strategy with losses for region and point tasks, and demonstrates superior performance over 13 baselines on SurgBlood and HemoSet, highlighting improved robustness in dynamic surgical environments. The combination of a real dataset, memory-aware temporal modeling, and prompt-based cross-task integration has practical implications for real-time surgical decision support and hemostasis management.

Abstract

Intraoperative bleeding in laparoscopic surgery causes rapid obscuration of the operative field to hinder the surgical process and increases the risk of postoperative complications. Intelligent detection of bleeding areas can quantify the blood loss to assist decision-making, while locating bleeding points helps surgeons quickly identify the source of bleeding and achieve hemostasis in time to improve surgical success rates. To fill the benchmark gap, we first construct a real-world laparoscopic surgical bleeding detection dataset, named SurgBlood, comprising 5,330 frames from 95 surgical video clips with bleeding region and point annotations. Accordingly, we develop a dual-task synergistic online detector called BlooDet, enabling simultaneous detection of bleeding regions and points in laparoscopic surgery. The baseline embraces a dual-branch bidirectional guid- ance design based on Segment Anything Model 2. The mask branch detects bleeding regions through adaptive edge and point prompt embeddings, while the point branch leverages mask memory to induce bleeding point memory modeling and captures point motion direction via inter-frame optical flow. By coupled bidirectional guidance, our framework explores spatial-temporal correlations while exploiting memory modeling to infer current bleeding status. Extensive experiments indicate that our method outperforms 13 counterparts in bleeding detection.

Figures (9)

• Figure 1: (a): Illustration of bleeding detection task with samples in SurgBlood and predictions of our solution. (b): The proposed BlooDet performs dual-branch bidirectional guidance for synergistic bleeding region and point detection.
• Figure 2: Illustration of bleeding types in SurgBlood.
• Figure 3: Statistical distribution of video clips in SurgBlood.
• Figure 4: Bleeding distribution. Left: proportion of frames with bleeding region and point; Right: Distance of bleeding region center and bleeding point to image center.
• Figure 5: Overview of the proposed BlooDet. Our framework comprises a mask branch and a point branch to jointly detect bleeding regions and bleeding points. Cross-branch guidance and adaptive prompt embedding allow our framework to reach a co-optimized state.