Development of Image Analysis Algorithms for Computational Assessment of Cardiac Structure and Function

在过去的十年中，心脏磁共振成像技术的进步为开发新的测量方法来检测心脏结构和功能的异常提供了巨大的潜力。这两项进步包括获得空间分辨率更高的高分辨率（Hi-res）三维（3D）图像，以及区分心肌损伤（梗死）组织和健康组织的定量成像技术。梗死区组织在导致心功能障碍的机制中起着核心作用，因此，准确和可重复的梗死区估计对心脏结构和功能的分析至关重要。与这些发展相平行，心功能的计算建模已经成为预测心脏异常电活动的一种很有前途的工具。这些计算模型是基于从心脏MR图像中提取的信息建立的，并且必须结合心室和心脏梗死区域的精确几何形状以进行准确预测。这为我们提供了一个机会来评估，与传统的低分辨率图像生成的模型相比，在构建计算模型时采用高分辨率定量成像方法是否会提高预测心脏异常电活动的灵敏度和特异性。然而，为了从这些三维磁共振图像中提取几何信息，需要开发鲁棒的自动化图像分析方法。该研究计划的总体目标是通过开发用于心脏结构及其功能的非侵入性评估的先进图像分析工具来解决这一需求。特别是，将开发图像分析方法，将心脏MR图像对齐到一个共同的参考，以纠正心脏运动并提取心脏的三维心室和梗死几何形状。心脏电活动的计算模拟将被用来评估3D磁共振技术的有效性。该研究项目将通过医学图像分析的新发展，整合心脏磁共振成像和计算心脏病学的最新进展。这项工作对生物医学工程和成像科学的预期意义是双重的：由3D磁共振图像的新图像分析方法的发展带来的工程创新；使我们能够理解准确的方法来成像梗塞结构。通过准确地结合梗死结构，这些虚拟电生理模型可以提供一种无创检查心脏的工具，而不需要侵入性的导管方法。虽然所提出的算法最初将用于心脏MR图像，但它们将直接适用于分析其他成像技术和器官的图像

Advancements in cardiac magnetic resonance (MR) imaging technologies in the past decade have enabled great potential for developing novel measurements to detect abnormalities in the structure and function of the heart. Two such advancements include acquisition of high-resolution (Hi-res) three-dimensional (3D) images with increased spatial resolution, and quantitative imaging techniques to differentiate injured (infarcted) tissue from healthy tissue in the cardiac muscle. Tissue in infarcted regions plays a central role in the mechanisms that lead to heart dysfunction and therefore, accurate and reproducible estimation of the infarct regions is paramount to the analysis of cardiac structure and function. Parallel to these developments, computational modeling of cardiac function has emerged as a promising tool to predict abnormal electrical activity of the heart. These computational models are built based on the information extracted from cardiac MR images and must incorporate accurate geometries of the ventricles and infarcted regions of the heart for accurate prediction. This presents us with an opportunity to evaluate whether incorporating Hi-res, quantitative imaging methods in building computational models, as compared to the models generated from conventional low-resolution images, will improve the sensitivity and specificity in predicting abnormal electrical activities of the heart. However, in order to extract geometric information from these 3D MR images, the development of robust, automated image analysis methods is required. The overall goal of the research program will be to address this need by developing advanced image analysis tools for the non-invasive assessment of the cardiac structure and its function. In particular, image analysis methods will be developed to align cardiac MR images to a common reference to correct for cardiac motion and extract the 3D ventricular and infarct geometry of the heart. Computational simulations of the electrical activity of the heart will then be used to evaluate efficacy of the 3D MR technique. This research program will integrate emerging advances in cardiac MR imaging and computational cardiology through novel developments in medical image analysis. The anticipated significance of this work to biomedical engineering and imaging science is twofold: engineering innovations resulting from the development of new image analysis methods for 3D MR images; and enabling our understanding of accurate methods to image the infarct structure. By accurately incorporating the infarct structure, these virtual electrophysiological models may provide a tool to non-invasively interrogate the heart without the need for invasive catheter-based methods. Although the proposed algorithms will initially be developed for cardiac MR images, they will be directly applicable to analyzing images of other imaging techniques and organs.**