Recent advances in multi-band excitation techniques have been remarkable. However, accurate slice separation from collapsed data that were simultaneously acquired at multiple slice locations continues to be challenging, especially with respect to the use of animal scanners that are equipped with relatively few receiver coil elements (i.e., four channels). The purpose of this study was to reduce EPI artifacts and achieve more accurate and robust separation of simultaneously acquired slices that are acquired in multi-band EPI by using an iterative reconstruction method.

Subjects and MRI scans: In vivo male Wister rat brains (n = 5; average body weight = 300 g each) were scanned on a 7T animal scanner (Bruker, BioSpec70/20) equipped with a transceiver RF coil and a receiver RF coil with four channel elements. A custom single-shot dual-band 2D EPI sequence was used with blipped-controlled aliasing (CAIPI) for the main scans; meanwhile, a single-shot single-band 2D EPI sequence was used for the reference scans. Dual-band excitation pulses were designed based on sync functions with frequency offsets. The between-band spacing was 10.33 mm. The slice acquisition order was an interleaved multi-slice acquisition (from caudal to rostral) without an inter-slice gap. A total of 62 slices (31 excitations) were imaged, in a single repetition, to cover the entire brain. The scan-timing parameters of repetition time (TR) / echo time (TE) = 1,700/14.6 ms. The voxel size in the plane was 0.25 × 0.25 mm, and the slice thickness was 0.333 mm. The encoding matrix sizes in the read-out and phase-encoding directions were 96 and 56, respectively. The in-plane imaging matrix size was 96 × 56, while the flip angle was 80°. Neither in-plane acceleration (GRAPPA, etc.) nor partial Fourier sampling was used.

Pre-scans: For phase corrections, pre-scans were performed that consisted of both single-band and dual-band EPI acquisitions, with or without partial phase-encoding blips; these included a set of scans without PE blips (0%), with half-sized PE blips (50%), and with full-sized PE blips (100%) (Fig. 1).

Reconstruction procedures: To separate two slices simultaneously acquired from different locations, a novel kernel method was performed, in which an iterative calculation of 1 × 1-sized kernels was used to estimate each point (i.e., point-by-point procedure) in the k-space. The conventional slice GRAPPA method (with 1 × 1 kernel size) was also performed, for comparison purposes. An iterative kernel method was applied to separate data simultaneously acquired from two slice locations, into two individual slice datasets; this was done according to information from the coil sensitivity profiles of multi-element array coils.

Iterative procedure for estimating and applying the kernels for slice separation: (see Fig.2)

Signal sensitivity correction between the two simultaneously acquired slices was performed using a combination of pre-scan data obtained by using both single-band and dual-band excitation RF pulses.

Phase corrections: In the image reconstruction process involving the dual-band acquisitions, EPI phase correction was performed after the slice separation processes were carried out. Using the EPI pre-scan data with various sizes of phase-encoding gradients, 2D phase difference maps as well as conventional 1D phase differencing along the read-out direction were calculated for phase correction in the main scans.

The proposed method was effective and valid in reducing artifacts in dual-band EPI in animal scanners. The results obtained using the proposed method were more robust and accurate than those obtained by using the conventional method (Figs. 3). Dual-band RF pulses and single-band RF pulses might differ slightly in terms of sensitivity profile, especially in the areas around the coil element boundary (i.e., the boundary between the sensitivity profiles of the two adjacent elements of the RF coil). The results indicate that the proposed kernel computation method could minimize errors in the estimation of coil sensitivity profiles between the reference scans (i.e., with single-band excitation) and the main scans (i.e., with dual-band excitation).A method was proposed to reduce artifacts in multi-band EPI; this iterative method could accurately separate collapsed k-space data acquired simultaneously from multiple slice locations. Our findings give rise to a more efficient multi-band EPI technique that can be used even with a scanner system equipped with relatively few receiver coil elements. The use of this method would enable faster acquisition of EPI results, with the efficient whole-brain coverage of small animal brains—all without sacrificing spatial resolution, especially in the slice-encoding direction.