4D-CTA 数据集
收藏arXiv2025-02-14 更新2025-02-27 收录
下载链接:
http://arxiv.org/abs/2502.09893v1
下载链接
链接失效反馈官方服务:
资源简介:
本研究构建了一个名为4D-CTA的数据集,该数据集由丹娜-法伯癌症研究院提供,包含了25位患者的头部分辨率减影CT图像。这些图像用于创建颅内动脉和静脉的血管模板,旨在帮助提高脑血管疾病的诊断效率。数据集通过4D-CTA技术获取,包含了动脉和静脉不同阶段的图像,用于训练和验证深度学习模型,以实现血管的自动分割。
A dataset named 4D-CTA is constructed in this study, which is provided by Dana-Farber Cancer Institute. The dataset contains high-resolution subtraction CT head images from 25 patients. These images are utilized to create vascular templates for intracranial arteries and veins, aiming to improve the diagnostic efficiency of cerebrovascular diseases. Acquired via 4D-CTA technology, the dataset includes images of different phases for both arteries and veins, and is used to train and validate deep learning models for automatic vascular segmentation.
提供机构:
丹娜-法伯癌症研究院,波士顿,MA
创建时间:
2025-02-14
搜集汇总
数据集介绍

构建方式
4D-CTA 数据集的构建基于动态计算机断层扫描血管造影技术,旨在通过捕获颅内血管的时相信息,为脑血管疾病的诊断提供关键支持。数据集从机构研究数据库中检索了头部 4D-CT 数据,并从增强后的图像中减去了骨骼和软组织。通过高级归一化工具(Advanced Normalization Tools)流程,从 25 名患者中创建了血管造影模板。随后,使用非线性配准技术,通过 CT衰减阈值确定了由模板驱动的感兴趣区域(ROIs),以生成动脉和静脉的分割。此外,利用 MRA 血管分割工具 iCafe 在 29 名患者的动脉和静脉结构上进行了分割,并以此训练了深度学习(DL)模型,以骨骼在内的 CT 图像作为输入。4D-CT 中的多个时相图像被用于增加训练和验证数据集。
特点
该数据集的特点在于其动态性和高时间分辨率,能够捕捉颅内血管的时相变化,从而更准确地反映血管的真实状态。此外,数据集通过预处理和分割技术,成功地将骨骼和软组织从增强后的图像中去除,突出了血管结构,为血管分割和模板构建提供了良好的基础。最后,该数据集还包含了通过深度学习模型和模板驱动的分割方法获得的分割结果,以及由神经放射科医生验证的真实分割结果,为评估分割技术的准确性提供了宝贵的参考。
使用方法
4D-CTA 数据集可用于开发和应用各种图像分析技术,以辅助专家对颅内 CTA 进行解释。例如,可以基于该数据集开发血管模板,以理解正常的血管形态,或自动分割 Circle of Willis 以评估血管架构。此外,还可以使用该数据集训练深度学习模型,以自动分割 CTA 中的动脉血管,从而减轻放射科医生的认知负担。最后,该数据集还可用于评估不同分割技术的准确性,并为进一步改进和优化分割技术提供数据支持。
背景与挑战
背景概述
computed tomography angiography (CTA) is a vital tool in the diagnosis of cerebrovascular diseases. It captures detailed images of blood vessels in the brain, which is crucial for detecting abnormalities such as stenoses and aneurysms. Dynamic CTA, a type of imaging that captures temporal information about the contrast passage, has been instrumental in advancing this field. The 4D-CTA dataset, developed by researchers at Dana-Farber Cancer Institute, Brigham and Women’s Hospital, and Harvard Medical School, represents a significant contribution to the field. This dataset was created to address the lack of population-averaged vessel atlases for CTA and to develop and evaluate segmentation techniques. The research team aimed to create and register population-averaged vessel atlases and to employ deep learning for direct segmentation of vessels on CTA images. The dataset has been influential in the field, providing a valuable resource for the development of image analysis techniques that assist expert radiologists in interpreting CTA images for conditions such as strokes and aneurysms.
当前挑战
The 4D-CTA dataset presents several challenges that are intrinsic to the nature of CTA imaging and the segmentation process. The primary challenge lies in the creation of vessel templates due to the overlapping image intensities between contrast-enhanced vessels and the background, which is compounded by the intricate anatomy of the cerebral vasculature. The dataset also addresses the challenge of vessel segmentation in CTA images, which lag behind those developed for Magnetic Resonance Angiography (MRA) due to the lower contrast-to-background ratio. Additionally, the extraction of ground truth segmentations is a time-consuming and expensive task, requiring expert annotation. The 4D-CTA dataset uses dynamic imaging and subtraction techniques to mitigate some of these challenges, but the limited size of the dataset and the need for advanced CT scanners to achieve optimal results present further limitations. Despite these challenges, the dataset has been instrumental in demonstrating the potential of dynamic CTA and deep learning models for accurate and robust vessel segmentation across varying contrast phases, which has significant implications for enhancing clinical decision-making and patient care in the diagnosis of cerebrovascular diseases.
常用场景
经典使用场景
4D-CTA 数据集广泛应用于脑部血管疾病的诊断中。其动态成像能力能够捕捉血管随时间变化的特性,对于检测诸如血管狭窄和动脉瘤等异常情况具有重要意义。该数据集通过创建和注册人群平均血管图谱以及使用深度学习技术,实现了对血管的直接分割,极大地提高了诊断的效率和准确性。
实际应用
4D-CTA 数据集在实际应用中展现出巨大的潜力。在临床诊断中,该数据集可以帮助医生快速准确地检测脑部血管疾病,如中风和动脉瘤。此外,该数据集还可以用于血管病理学的深入研究,为开发新的治疗方法提供科学依据。同时,该数据集还可以用于训练深度学习模型,提高模型的泛化能力和鲁棒性,从而推动医疗影像分析技术的进一步发展。
衍生相关工作
4D-CTA 数据集的发布,衍生出了一系列相关的经典工作。这些工作包括基于该数据集的血管分割模型的开发、血管图谱的构建以及深度学习技术在医疗影像分析中的应用等。这些研究不仅推动了医疗影像分析技术的发展,也为临床诊断和治疗提供了有力的支持。
以上内容由遇见数据集搜集并总结生成



