Purpose Preventing intraoperative bleeding in endoscopic submucosal dissection (ESD) relies on the timely identification and pre-coagulation of submucosal vessels. We aim to reduce bleeding risk by delivering accurate vessel segmentation with uncertainty awareness. Methods We present ESD-VesNet, a prompt-free SAM3-based framework with an evidential head that outputs vessel probability and pixel-wise uncertainty. The model is trained in a false-positive-aware manner using our developed ESD-Vessel dataset, which includes 2401 positives (images with clear vessel annotation) and 708 curated hard negatives (images may be confused with vessels), combined with online uncertainty-guided hard negative mining. The model is trained in a false-positive-aware manner using the ESD-Vessel dataset, which includes 2401 positive samples (images with clear vessel annotations) and 708 curated hard negatives (images visually confusable with vessels), combined with online uncertainty-guided hard negative mining. By suppressing false positive predictions, our approach has the potential to minimize unnecessary interventions, such as unwarranted coagulation. Results On the validation set, ESD-VesNet achieves Dice<InlineEquation ID="IEq1"> <EquationSource Format="TEX">$_{FG}$</EquationSource> <EquationSource Format="MATHML"><math> <mmultiscripts> <mrow /> <mrow> <mi mathvariant="italic">FG</mi> </mrow> <mrow /> </mmultiscripts> </math></EquationSource> </InlineEquation> 52.11%, IoU<InlineEquation ID="IEq2"> <EquationSource Format="TEX">$_{FG}$</EquationSource> <EquationSource Format="MATHML"><math> <mmultiscripts> <mrow /> <mrow> <mi mathvariant="italic">FG</mi> </mrow> <mrow /> </mmultiscripts> </math></EquationSource> </InlineEquation> 35.23%, mDice 75.70%, and Vessel Detection Rate 96.81%, while keeping Background False Positive Rate at 0.61%. Structural quality is high (E-measure 0.7275; S-measure 0.7223). Qualitative analyses show robustness under bubbles, debris, reflections, and blood seepage. Conclusion Integrating SAM3 with evidential uncertainty and hard negative mining yields a clinically oriented vessel segmentation system that maintains high detection sensitivity with low false positives, supporting safer ESD.

ESD-VesNet: uncertainty-aware vessel segmentation network for endoscopic submucosal dissection with hard negative mining

期刊论文
出版模式：

开放获取
发表日期：

2026年7月3日

查看全文

Mengya Xu,
Ming Chen,
Zhen Li,
Chaoyang Lyu,
An Wang,
Rulin Zhou,
Chuanhao Zhao,
Jiaxun Xiang,
Tsz Chun Wong,
Hossein Farahnaki,
Sobhan Zamani Kiasari,
Tong Wu,
Zimeng Su,
Yile Zeng,
Ruijing Wen,
Xiaohan Shang,
Yi Mu,
Kezhen Lin,
Yidong Zhang,
Hongliang Ren

摘要

Purpose

Preventing intraoperative bleeding in endoscopic submucosal dissection (ESD) relies on the timely identification and pre-coagulation of submucosal vessels. We aim to reduce bleeding risk by delivering accurate vessel segmentation with uncertainty awareness.

Methods

We present ESD-VesNet, a prompt-free SAM3-based framework with an evidential head that outputs vessel probability and pixel-wise uncertainty. The model is trained in a false-positive-aware manner using our developed ESD-Vessel dataset, which includes 2401 positives (images with clear vessel annotation) and 708 curated hard negatives (images may be confused with vessels), combined with online uncertainty-guided hard negative mining. The model is trained in a false-positive-aware manner using the ESD-Vessel dataset, which includes 2401 positive samples (images with clear vessel annotations) and 708 curated hard negatives (images visually confusable with vessels), combined with online uncertainty-guided hard negative mining. By suppressing false positive predictions, our approach has the potential to minimize unnecessary interventions, such as unwarranted coagulation.

Results

On the validation set, ESD-VesNet achieves Dice $_{FG}$ $_{FG}$ 52.11%, IoU $_{FG}$ $_{FG}$ 35.23%, mDice 75.70%, and Vessel Detection Rate 96.81%, while keeping Background False Positive Rate at 0.61%. Structural quality is high (E-measure 0.7275; S-measure 0.7223). Qualitative analyses show robustness under bubbles, debris, reflections, and blood seepage.

Conclusion

Integrating SAM3 with evidential uncertainty and hard negative mining yields a clinically oriented vessel segmentation system that maintains high detection sensitivity with low false positives, supporting safer ESD.