Image-based ventricular blood flow analysis

We describe a novel and non-invasive method for the quantitative analysis of blood flow in the left ventricle of specific human patients using information derived from medical images. There are three major components to the method. First, a new approach to the segmentation problem is presented which locates the ventricle (or other organs of interest) in medical images. The method is independent of the imaging modality used (for example, MR, CT, and Ultrasound) and is automatic, requiring as initialization a single point within the interior of the ventricle. Existing segmentation techniques either require much more information during initialization, such as an approximation to the boundary of an object, or are not robust to the types of noisy data encountered in the medical domain. By integrating region-based and physics-based modeling techniques we have devised a hybrid design that overcomes these limitations. In our experiments we demonstrate, across imaging modalities, that this integration automates and significantly improves the boundary detection results. Next, a technique known as MRI-SPAMM is applied to extract the full 3D motion of the ventricular walls. This technique applies a magnetic grid within the heart tissue which deforms along with the tissue through the cardiac cycle and appears in MR images. By tracking the deformation in the images using a physics-based model, the wall motion can be quantified. Finally, the ventricular wall motion information is used by an efficient computational fluid dynamics solver to simulate the flow of blood through the ventricle. Boundary conditions for the solver are directly derived from the wall motion information, which allows for the first time a patient-specific LV blood flow simulation. We present experiments using data from both normal and diseased subjects and compare our results with other techniques for estimating ventricular blood flow.