The accurate quantification of transmural left ventricular (LV) regional function is crucial for managing patients with ischemic heart disease. Previous imaging-based efforts have been hampered by the limitations of conventional two-dimensional imaging and inadequate image analysis methodology (e.g., the inability to handle out-of-plane motion). More recently, these limitations are starting to be overcome by three- dimensional (3D) imaging strategies and methods to estimate pointwise myocardial displacements and/or velocities. However, the task of assembling dense 3D sets of image-derived displacement/velocity data into transmural, quantitative measurements of LV function remains a difficult one. The research proposed here is aimed at further development of an image analysis strategy to more accurately, robustly and reproducibly characterize 3D LV function over time (i.e., 4D datasets) by quantifying the transmural deformation of the LV from any one of several imaging modalities. At the core of the effort is an approach that uses both kinematic and biomechanical Finite Element models, computer vision-related strategies based on differential geometric features and mathematical optimization reasoning methods to estimate the non-linear, non-rigid deformation of the left ventricle of the heart. A key feature of this approach is that it can combine surface displacement information at the endocardial and epicardial boundaries derived from the geometric, shape- based tracking strategies developed in the first three years of this grant with midwall velocity information, such as that found from MRI phase velocity data, echocardiographic Doppler tissue velocity data or velocity data estimated using a physical flow approach to form 3D maps of transmural myocardial strain estimates which can be broken down into directional components. The accuracy and reproducibility of the approach, as well as the importance of using multiple sources of displacement/velocity information, will be demonstrated by validating the image-derived strain measurements obtained from a range of imaging modalities (cine gradient echo and echo planar MRI, 3D echocardiography (3DE) and 3D cine Computed Tomography from the Dynamic Spatial Reconstructor (DSR)) to those made using 1.) well-characterized material samples and ii.) arrays of image-visible markers implanted in acute dogs. The utility of the approach for deriving clinically-relevant measurements related to the transmurality of infarct morphology will be tested by comparing MRI-derived measurements of strain in two groups of acute dogs having different post mortem-determined degrees of transmural and non- transmural infarction. Finally, we will test the algorithm on a number of 4D image datasets of normal human subjects in each of two modalities (MRI and 3DE) to establish clinical feasibility and reproducibility.
