Black-blood cardiac MRI can be done either via the use of the outflow effect or with a double inversion recovery (DBIR) preparation. The former allows whole heart coverage with single-shot 2D acquisitions but come at the cost of poor flow suppression and low spatial resolution. A DBIR fast spin-echo (FSE) sequence can generate black-blood images with good flow suppression and high spatial resolution but is limited to being a single slice technique due to the non-selective inversion pulse used in the DBIR module. Techniques to increase slice coverage in DBIR-FSE tradeoff SNR by blood nulling quality [1,2,3]. Multi-band (MB) excitation allows better SNR efficiency (as the slices are excited and imaged simultaneously) and the slices are acquired at the same inversion time [4]. In this work we present a technique that supports multi-slice coverage in cardiac triggered breath-held black-blood acquisitions with MB excitation and present a framework to obtain accurate T2 maps from all the slices.

MB pulses for 4 slices were generated by a summing design. The pulses were designed to spatially encode the signal in the slice dimension according to the following matrix. $$\begin{bmatrix}-1& 1& 1& 1 \\1& -1& 1& 1 \\1& 1& -1& 1 \\1& 1& 1& -1 \\\end{bmatrix} $$ Four slices are excited simultaneously (covering a 27.5 mm volume with 5 mm slices) and a slab selective refocusing pulse was used to refocus a 33 mm slab. The selective inversion pulse in the DBIR module inverts the entire 33 mm slab. The MB excite pulses were used with a radial DBIR-FSE sequence as shown in Fig. 1. The k-space data from a single acquisition can be separated into individual TEs which are reconstructed to generate high-resolution TE images and T2 maps. Imperfections in the refocusing pulses leads to indirect echoes which cause a deviation of the signal decay from a single exponential and this problem is amplified in MB excitation. Figure 2a shows the individual slice profiles that are simultaneously excited by the MB excitation pulse along with the refocusing pulse profile. It can be noticed that with non-rectangular refocusing profiles the edge slices are more contaminated by indirect echoes such as stimulated echoes. This results in a signal evolution that is different from the two slices in the center of the slab Fig 2b. Using a simple mono-exponential signal model to estimate the T2s will result in estimation errors. Recently, a slice-resolved extended phase graph (SEPG) fitting algorithm was proposed [6] to accurately estimate T2s from data contaminated by indirect echoes and this model was incorporated into an iterative algorithm (CURLIE) to reconstruct TE images from under-sampled radial FSE data [5] by solving the following optimization problem $$argmin_{I_0, T2, B1} \sum_{j=1}^{n} || FT[C_j(I_0, T2, B1, \alpha_0,\ldots,\alpha_j)] - K_j ||_2^2 $$ In the above equation FT is the forward Fourier Transform, is the undersampled k-space data at the jth TE, and $$$C_j$$$ is the SEPG model expressed as a function of $$$I_0$$$, T2, B1 and the profiles of the excitation and refocusing RF pulses. The non-linear SEPG signal model is linearized by using a principal component approach and solved using a conjugate gradient algorithm [5].

To use the SEPG model, the excitation profile of the individual slices is used along with the refocusing slice profile to generate the training SEPG signal decays used to generate the principal components. The T2 maps for each slice are then estimated by fitting the TE decay curves per pixel using the SEPG model.

Short axis data were acquired at 3T (Skyra, Siemens) with the radial MB-DBIR-FSE pulse sequence on normal volunteers. A total of 96 views were acquired for each MB encode with ETL=16, echo spacing=8.1 ms, bandwidth=501 Hz/pixel and TR=1 R-R. The acquisition interleaves the MB pulses across TR’s.

Data acquired with the MB-DBIR-FSE pulse sequence were reconstructed to yield anatomical black-blood images, 16 TE images (reconstructed from 6 radial lines per TE) and T2 maps for each of the four slices. The 64 images and corresponding T2 maps are obtained from data acquired in a single breath hold. The 4 slice black-blood images and the corresponding T2 maps are shown in Fig. 3.A technique that can generate several black-blood images and the associated T2 maps in the myocardium is presented. Due to the high acceleration factors and the substantial deviation of the refocussing pulse from ideal behavior, a model based scheme which accounts for this non-ideal behavior was used in conjunction with sparsity constrained reconstructions to yield T2 maps from all the slices.