Strategies for Generating Multi-Time Frame Localization of Cardiac MRI

Abstract

4D Flow MRI is a recent promising technology that is able to capture blood flow information within the heart chambers over a cardiac cycle. To accurately study the flow inside the chambers, there is a need for a high quality anatomical reference which can be provided by another scan known as 3D cine MRI (short-axis 3D (multiple 2D slices) cine SSFP). To take advantage of both scans, data fusion can be done using an intensity-based registration. To reduce the impact of noise on the registration result and the chance of misalignment between the organs, defining a region of interest (localization) should be done prior to the registration. Localizing a dataset - especially a time-varying dataset - can be a daunting task since the localization should be provided for all time frames. We design and evaluate different strategies for extending single time frame localization to time varying data in order to register the 4D Flow MRI and 3D cine MRI over the cardiac cycle.