---
title: "Flow Matching Meets 3D Curvilinear Structure Segmentation in Medical Imaging"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.19965v1"
permalink: "https://feed.astralor.com/items/itm_3088f2bd42dc7b7a045b1f4d"
published: "2026-08-20T12:35:27.000Z"
score: 75
channel: "tech"
tags: ["3D-CurvSegFlow","Flow Matching","Medical Image Segmentation","Diffusion Models"]
---

# Flow Matching Meets 3D Curvilinear Structure Segmentation in Medical Imaging

## AI commentary

> If your team has been deterred by diffusion models' compute cost for 3D segmentation, this paper is a must-read: flow matching achieves comparable or better structure preservation with far fewer steps. For medical imaging engineers, 3D-CurvSegFlow offers a plug-and-play general framework that avoids per-organ model tuning.

## At a glance

Segmentation of curvilinear anatomical structures in 3D medical images remains challenging due to complex topology, severe class imbalance, weak contrast, and large variations in structure morphology. Existing deep learning methods are often tailored to specific anatomies or modalities, limiting generalization; diffusion-based methods suffer from expensive sampling. This paper presents 3D-CurvSegFlow, a flow matching-based model that learns a continuous transformation from a simple source distribution to the target vascular representation, enabling progressive refinement with efficient inference. Evaluated on three public datasets covering portal vein, cerebral vessels, and coronary arteries, the method outperforms general-purpose and vessel-specific approaches using a common architecture and training strategy, with strong preservation of thin branches and vascular continuity. This work advances state-of-the-art in 3D curvilinear segmentation and opens avenues for efficient, generalizable, and clinically applicable medical image analysis.

## Key points

- Introduces 3D-CurvSegFlow, the first application of flow matching to 3D curvilinear structure segmentation, replacing costly diffusion models.
- Outperforms general-purpose and vessel-specific approaches with a unified architecture and training strategy across three public datasets.
- Preserves thin branches and vascular continuity effectively, with more efficient inference than diffusion models for high-resolution 3D volumes.

## Full text

Segmentation of curvilinear anatomical structures in 3D medical images remains challenging due to complex topology, severe class imbalance, weak contrast, and large variations in structure morphology. While deep learning approaches for 3D curvilinear segmentation have been proposed, they are often tailored to specific anatomies or modalities, limiting generalization across clinical settings and leaving room for improvement. Recent generative models have shown the benefits of iterative prediction for structured segmentation tasks, yet diffusion-based methods suffer from computationally expensive sampling, hindering their use on high-resolution 3D volumes. We present 3D-CurvSegFlow, a flow matching-based model for 3D curvilinear structure segmentation. The model learns a continuous transformation from a simple source distribution to the target vascular representation, enabling progressive refinement of complex curvilinear geometries with efficient inference. We evaluate our method on Three public challenging datasets covering distinct anatomies and modalities: portal vein, cerebral vessel, and coronary arteries. Using a common architecture and training strategy across all tasks, our method outperforms general-purpose and vessel-specific approaches, with strong preservation of thin branches and vascular continuity. This work not only advances the state-of-the-art in 3D curvilinear segmentation but also opens new avenues for efficient, generalizable, and clinically applicable methods in medical image analysis.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_3088f2bd42dc7b7a045b1f4d