An image-based phase correction method for multi-band Echo Planar Imaging (EPI), in conjunction with the slice separation technique using a point-by-point procedure in k-space, was proposed in this study. To achieve the EPI with dual-band excitation in animal scanners equipped with relatively few coil elements (i.e., 4 ch), we need a robust correction in the phase difference between the odd and even phase encoding lines, as well as accurate techniques for the multiband acquisition and slice separation. This study aimed to show with the actual MRI measurement that the proposed image-based phase correction can be applied to and is useful for the dual-band EPI, if it is used with the point-by-point procedure in k-space (1 × 1-sized kernel) in the slice separation procedure in reconstruction.

In vivo rat brains were scanned on a 7T animal scanner (BioSpec 70/20, Bruker) equipped with a 4-channel receiver coil. A single-shot dual-band gradient echo 2D-EPI sequence with controlled aliasing (CAIPI) technique was used. The EPI pre-scans with various phase encoding blips (zero-, half-, and full-sized) were used as references to test the various types of phase correction methods for comparison. To separate the simultaneously acquired two slices from different locations, the slice GRAPPA method, except for the kernel size of 1 × 1 (conventional kernel size: 3 × 3), was used according to the information on coil sensitivity profiles of multi-element array coils. Image-based phase correction: In combination with point-by-point procedures in k-space (i.e., 1 × 1-sized kernel applications), image-based phase correction was applied to the EPI data with dual-band excitation. The first step in the proposed image-based phase correction method was to separate the odd and even lines in k-space. Fourier transformation was applied to them, and the odd and even images aliased along the phase encoding direction was obtained. Then, the non-overlapping (non-overlapped) regions in the aliased images were identified automatically from the pre-scan images obtained with half-sized phase encoding gradients. The phase difference map was generated from the odd and even images. Only the non-overlapped areas were masked and used for the fitting with 1D or 2D modeling to calculate the phase difference.

To achieve accurate phase correction, the proposed method was applied just once immediately after the slice separation procedures for the multi-band data. The results showed the usefulness and robustness of the proposed method compared with the conventional approach (Figs. 4, 5) because the pre-scan data do not always predict the main scan data. An image-based phase correction method is thus needed. However, implementing the image-based approach is difficult, especially in the case of scans with full-sized phase-encoding blips, because of their image overlapping by aliasing. Thus, we utilized the pre-scans with half-sized phase-encoding gradients as assistance.

Image ghosting because of erroneous phase correction is likely to be observed in animal MRI scanners with a high-performance gradient system (high slew rate and maximum gradient strength) because of the eddy current in the gradient coil circuits. The image-based approach and not the pre-scan-based one is therefore necessary for the robust phase correction. In conventional reconstruction steps using the slice-GRAPPA kernel with a 3 × 3-sized kernel for multi-band EPI data, phase correction is needed before the kernel application. This process essentially requires pre-scan-based 1D phase corrections. In this situation, using image-based phase correction techniques is difficult because the collapsed images acquired with the use of dual-band excitation with CAIPI (pi-shift) have the Nyquist ghost-like image from the other slice location shifted by one-half of the field-of-view (FoV) along the phase-encoding direction. By contrast, the proposed kernel method with a kernel size of 1 × 1 does not require the phase correction beforehand. After the slice separation from the collapsed data, phase correction was applied just once to the separated individual slice data. In the proposed approach, the image-based phase correction therefore becomes much easier to perform because it can be treated in the same way as in the case of the usual single-band EPI.

The proposed image-based phase correction method, in conjunction with the slice separation method using a point-by-point procedure in k-space, can be applied to and is useful in real dual-band EPI measurements of phantom and rat brains. The method should be validated in the application for functional, diffusion, and perfusion MRI. We have plans to test the proposed method on the MRI system for humans by using a receiver coil with many coil elements and involving large multi-band factors.