Segmentation of 3D Tubular Structures

计算机断层血管造影（CTA）正迅速成为排除冠状动脉疾病的诊断和指导工具。据预测，由于CTA的无创性和有希望的结果，CTA将减少30%的导尿次数。CTA使用的增加导致了大量的数据，但没有计算工具来对冠状动脉斑块进行分析。因此，迫切需要开发计算工具来自动检测冠状动脉，然后进行CTA动脉斑块的评估。为了评估如此大量的CTA数据，血管分割的过程需要自动化。这就需要在周围存在多个组织和造影剂分布不均匀的情况下，发展检测管状结构的方法。该项目的目标是为管状结构开发自动化的、数据驱动的特征检测，并开发基于特征的学习和预测算法，从而改进管状结构数据的分割，并将其应用于CTA领域。一个成功的结果有可能在提高医疗质量的同时降低成本。

Computer Tomography Angiography (CTA) is rapidly emerging as a diagnostic and guiding tool to rule out coronary artery disease in patients. It is predicted that, owing to its non-invasive nature and promising results, CTA will reduce the number of catheterizations by 30%. The increased use of CTA is resulting in a large amount of data, but there are no computational tools to perform analysis on coronary artery plaques. Hence, there is an urgent need to develop computational tools to automatically detect the coronary arteries and then proceed to the assessment of arterial plaque in CTA. To assess such a large amount of CTA data, the process of vessel segmentation needs to be automated. This will require the development of methods to detect tubular structures in the presence of multiple surrounding tissues and uneven distribution of contrast. The objectives of this project are to develop automated, data-driven feature detection for tubular structures, and to develop feature-based learning and prediction algorithms allowing an improved segmentation of tubular data and apply them to the domain of CTA. A successful outcome has the potential of improving health care quality while simultaneously reducing the cost.