Reconstruction of retrospectively-gated cardiac data using a combination of GRAPPA, SPACE-RIP, UNFOLD and an adaptive regularization scheme

Introduction: Retrospective gating is a preferred strategy for the reconstruction of cardiac cine data, as it proves superior to prospective gating for capturing end-diastolic phases. These phases are especially important for accurate ejection fraction measurements, as the left atrium contracts to increase the left ventricle’s blood volume. In the present work, desirable characteristics of UNFOLD [1], GRAPPA [2], SPACE-RIP [3] and kt-SENSE[4] have been carefully combined, leading to a fast and reliable cardiac cine reconstruction algorithm featuring very reasonable reconstruction times. The available acceleration was used here to increase the number of slices acquired per breath hold, to reduce exam duration. Two different approaches have been proposed to combine time-varying sampling schemes with retrospective gating [5, 6]. Both were implemented, found to be fairly equivalent in terms of image quality, and the approach from Ref.[5] proved about an order of magnitude faster in terms of processing speed. Both approaches are fully capable of accommodating any existing cardiac model for the uniform or non-uniform mapping of time samples onto the cardiac phase axis (non-uniform mappings account for the fact that diastole duration tends to be much more variable than systole duration in the presence of arrhythmia). In terms of parallel imaging reconstruction, regularization is critical to minimize noise amplification in cases where the inverse problem is ill-conditioned [7]. The advantageous scheme from kt-SENSE, based on prior-knowledge, was used here. The sampling scheme was similar to that adopted in SHRUG [8], with the central region undersampled by two-fold (Fig.1), for faster imaging than if the central region were fully sampled. This central region is used for sensitivity mapping purposes, and to provide prior knowledge for the regularization scheme.