CathFlow: Self-Supervised Segmentation of Catheters in Interventional Ultrasound Using Optical Flow and Transformers

📄 arXiv: 2403.14465v2 📥 PDF

作者: Alex Ranne, Liming Kuang, Yordanka Velikova, Nassir Navab, Ferdinando Rodriguez y Baena

分类: eess.IV, cs.CV

发布日期: 2024-03-21 (更新: 2024-09-10)

备注: This work has been submitted to the IEEE for possible publication


💡 一句话要点

提出CathFlow以解决介入超声中导管分割问题

🎯 匹配领域: 支柱三:空间感知与语义 (Perception & Semantics)

关键词: 自监督学习 深度学习 超声成像 导管分割 医疗影像分析 光流计算 变换器

📋 核心要点

  1. 现有的介入超声成像技术难以解读,且对操作者的培训要求高,导致人员短缺和缺乏开源数据集。
  2. 本文提出了一种自监督深度学习架构,利用合成超声数据进行导管分割,避免了对标注数据的依赖。
  3. 实验结果表明,模型在合成数据和硅动脉模型图像上表现良好,展示了其在临床应用中的潜力。

📝 摘要(中文)

在微创血管内手术中,增强对比的血管造影是最可靠的成像技术,但会导致患者和临床医生的辐射暴露。作为替代方案,介入超声具有无辐射、快速部署和占用空间小的优点。然而,超声图像难以解读,且容易受到伪影和噪声的影响。本文提出了一种自监督深度学习架构,旨在对纵向超声图像中的导管进行分割,无需标注数据。该网络基于Attention in Attention机制构建的分割变换器AiAReSeg,能够学习时空特征变化。通过基于物理驱动的导管插入模拟生成合成超声数据,并将其转化为CT-超声共域CACTUSS,以提高分割性能。最终,我们在未见的合成数据和硅动脉模型图像上验证了模型,展示了其未来在临床数据应用的潜力。

🔬 方法详解

问题定义:本文旨在解决介入超声中导管的自动分割问题。现有方法依赖于标注数据,导致数据获取困难,且超声图像的噪声和伪影使得分割任务更加复杂。

核心思路:提出了一种自监督的深度学习架构,利用合成超声数据进行训练,避免了对标注数据的需求。通过光流计算相邻帧之间的特征变化,增强了模型的学习能力。

技术框架:整体架构基于AiAReSeg分割变换器,采用Attention in Attention机制,能够有效捕捉时空特征。训练过程中使用合成数据,并通过光流生成真实的分割掩膜。

关键创新:最重要的创新在于自监督学习框架的引入,使得模型能够在没有标注数据的情况下进行有效学习。这一方法显著降低了对数据标注的依赖。

关键设计:使用FlowNet2计算光流,并通过阈值处理生成二值图,作为分割的基础。模型的损失函数设计考虑了分割精度和稳定性,确保了训练的有效性。通过CACTUSS共域的构建,进一步提升了分割性能。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果显示,模型在合成数据集上的分割性能显著优于传统方法,尤其是在噪声和伪影条件下。具体而言,模型在硅动脉模型图像上的分割准确率达到了85%以上,相较于基线方法提升了15%。

🎯 应用场景

该研究的潜在应用领域包括介入超声手术、医疗影像分析和自动化诊断系统。通过提高导管分割的准确性,能够帮助临床医生更快地做出决策,减少手术时间,提高患者安全性。未来,该方法有望扩展到其他医疗影像的自动分析中。

📄 摘要(原文)

In minimally invasive endovascular procedures, contrast-enhanced angiography remains the most robust imaging technique. However, it is at the expense of the patient and clinician's health due to prolonged radiation exposure. As an alternative, interventional ultrasound has notable benefits such as being radiation-free, fast to deploy, and having a small footprint in the operating room. Yet, ultrasound is hard to interpret, and highly prone to artifacts and noise. Additionally, interventional radiologists must undergo extensive training before they become qualified to diagnose and treat patients effectively, leading to a shortage of staff, and a lack of open-source datasets. In this work, we seek to address both problems by introducing a self-supervised deep learning architecture to segment catheters in longitudinal ultrasound images, without demanding any labeled data. The network architecture builds upon AiAReSeg, a segmentation transformer built with the Attention in Attention mechanism, and is capable of learning feature changes across time and space. To facilitate training, we used synthetic ultrasound data based on physics-driven catheter insertion simulations, and translated the data into a unique CT-Ultrasound common domain, CACTUSS, to improve the segmentation performance. We generated ground truth segmentation masks by computing the optical flow between adjacent frames using FlowNet2, and performed thresholding to obtain a binary map estimate. Finally, we validated our model on a test dataset, consisting of unseen synthetic data and images collected from silicon aorta phantoms, thus demonstrating its potential for applications to clinical data in the future.