Three-dimensional echo-shifted EPI with simultaneous blip-up and blip-down acquisitions for correcting geometric distortion

Abstract

Purpose: EPI with blip-up/down acquisition (BUDA) can provide high-quality images with minimal distortions by using two readout trains with opposing phase-encoding gradients. Because of the need for two separate acquisitions, BUDA doubles the scan time and degrades the temporal resolution when compared to single-shot EPI, presenting a major challenge for many applications, particularly fMRI. This study aims at overcoming this challenge by developing an echo-shifted EPI BUDA (esEPI-BUDA) technique to acquire both blip-up and blip-down datasets in a single shot.

Methods: A 3D esEPI-BUDA pulse sequence was designed by using an echo-shifting strategy to produce two EPI readout trains. These readout trains produced a pair of k-space datasets whose k-space trajectories were interleaved with opposite phase-encoding gradient directions. The two k-space datasets were separately reconstructed using a 3D SENSE algorithm, from which time-resolved B₀ -field maps were derived using TOPUP in FSL and then input into a forward model of joint parallel imaging reconstruction to correct for geometric distortion. In addition, Hankel structured low-rank constraint was incorporated into the reconstruction framework to improve image quality by mitigating the phase errors between the two interleaved k-space datasets.

Results: The 3D esEPI-BUDA technique was demonstrated in a phantom and an fMRI study on healthy human subjects. Geometric distortions were effectively corrected in both phantom and human brain images. In the fMRI study, the visual activation volumes and their BOLD responses were comparable to those from conventional 3D echo-planar images.

Conclusion: The improved imaging efficiency and dynamic distortion correction capability afforded by 3D esEPI-BUDA are expected to benefit many EPI applications.

Keywords: 3D EPI; Echo shifting; Hankel low-rank reconstruction; blip-up and blip-down acquisitions (BUDA); fMRI; geometric distortion correction.

Figure 1:

A schematic to illustrate 3D echo-shifted EPI with blip-up and blip-down acquisitions (esEPI-BUDA) in a single TR (A), and the corresponding two k-space trajectories (B). Echo-shifting gradients (shaded in blue) are applied along the slab-selection direction ($G_z$) to select the signals for the two echo-train acquisitions (red and green) resulting from the first and second RF pulses, respectively. A small gradient with ½ phase-encoding blip area ($G_y$) is played out prior to the second echo-train so that the two k-space trajectories are interleaved as shown in (B). The sequence can be readily modified for 2D multi-slice imaging by changing the slab-selection gradient to slice-selection gradient while eliminating the stepping phase-encoding gradient in $G_z$.

Figure 2:

Illustration of the image reconstruction steps involved in esEPI-BUDA by using a section of the human brain image as an example. Each under-sampled echo-train dataset (i.e., Echo-train 1 with blip-up acquisition and Echo-train 2 with blip-down acquisition, as shown in Figure 1B) first underwent 3D SENSE reconstruction individually, followed by TOPUP in FSL to estimate a B₀-field map. The B₀-field map was subsequently incorporated to jointly reconstruct the data from both echo-trains with Hankel structured low-rank regularization for geometric distortion correction.

Figure 3:

Representative images of a section selected from the 3D datasets of the DQA phantom acquired using 3D EPI with separate blip-up (A) and blip-down (B) acquisitions, the corresponding 3D EPI TOPUP image (C), individually SENSE-reconstructed images from the first (D) and second (E) echo train of the 3D esEPI-BUDA sequence, the resultant esEPI-BUDA image (F) with joint reconstruction, and a conventional 3D SPGR image (G). Image distortion was substantially corrected in (C) by using TOPUP in FSL, and in (F) by using esEPI-BUDA which jointly reconstructed an image from two interleaved k-space datasets with reversed k-space trajectories in one single shot. Using the distortion-free SPGR image as a benchmark, the esEPI-BUDA image in (F) demonstrated a greater similarity with a SSIM of 0.91 and a NRMSE of 0.06 than the EPI TOPUP image which yielded an SSIM of 0.87 and an NRMSE of 0.08.

Figure 4:

Representative whole-brain 3D images from a healthy human subject. (A) and (B): SENSE-reconstructed images using the first and the second echo train in Figure 1A, respectively; (C): the corresponding 3D esEPI-BUDA images from both echo trains acquired in a single scan.

Figure 5:

Representative images of a slice with a voxel size of 3.1 × 3.1 × 4 mm³ selected from the 3D datasets of the human brain. (A) and (B) were acquired using 3D EPI with separate blip-up (A) and blip-down (B) acquisitions. (C) displays an image reconstructed from (A) and (B) by using 3D EPI TOPUP. (D) and (E) were reconstructed from the first and second echo train of the esEPI-BUDA sequence, respectively. (F) shows the resultant esEPI-BUDA image reconstructed from both echo trains. (G) displays a conventional 3D SPGR image. Image distortion was effectively corrected in (C) and (F).

Figure 6:

A set of 120 B₀-field maps (A) in a slice arbitrarily selected from the 3D datasets of the human brain, covering a total time span of 4 min and 48 sec with a temporal resolution of 2.4 sec. The dynamic B₀ evolutions at the frontal (blue box) and occipital (orange box) lobes are displayed in (B). The corresponding spatial dislocation caused by B₀ are shown in (C), as calculated from Eq. (4). The maximal B₀ shifts during the process were approximately 6.2 Hz and 4.7 Hz in the two brain regions, corresponding to the spatial dislocations of 1.95 mm and 1.47 mm, respectively.

Figure 7:

Representative fMRI visual activation maps of 3D EPI with separate blip-up (A) and blip-down (B) acquisitions, the corresponding 3D EPI TOPUP (C), and 3D esEPI-BUDA (D), overlaid onto the T1-weighted MP-RAGE images. The activation maps in (D) showed the best spatial correspondence with the brain parenchyma in the T1-weighted structural images, when compared with the activation maps in (A) – (C). The color bar indicates the scale of t-values.

Figure 8:

The averaged time course with standard deviations across the six subjects from the 3D EPI with separate blip-up (A) and blip-down (B) acquisitions, the corresponding 3D EPI TOPUP (C), and the 3D esEPI-BUDA sequence (D). Both 3D esEPI-BUDA and conventional 3D EPI produced similar BOLD signal change (~3%), which was higher than 3D EPI TOPUP (~2%). The black bar in each sub-figure represents the visual stimulation time period, which was 24 sec.