Researchers and clinicians whose investigations require high resolution and high-quality Whole Body Diffusion-Weighted Imaging (WBDWI).

Whole Body Diffusion Weighted Imaging (WBDWI) is becoming increasingly relevant to detect, characterize and monitor cancerous tissue¹. Single-shot Echo Planar Imaging (ssEPI) is the standard pulse sequence for whole body diffusion acquisition. However, despite combining with parallel imaging acceleration, ssEPI still suffers severe shortcomings for high-resolution multi-station acquisition outside of the brain, due to concurrent constraints including increased susceptibility artifacts, physiological motion, impaired coil penetration and the demanding anatomical requirements for larger imaging field-of-view (FOV). In this study, we examine the feasibility of using an accelerated multishot acquisition reconstruction scheme to achieve high in-plane spatial resolution with improved image quality over an extended FOV. The algorithm was initially developed for high-resolution brain imaging and we report preliminary results for its application to WBDWI.

Data acquisition

A healthy volunteer (male, 35y.o) was scanned on a 1.5T MR450w scanner (GE Healthcare, Waukesha, WI, USA) using the body-coil array for transmission and a combination of the HNU coil with a 32-channel GEM body receiver-array for reception. A 4-stations acquisition was performed in 3 different ways: a) using conventional single-shot EPI DWI with following parameters: FOV:46cm(LR) x 36.8cm(AP), Matrix:96(freq) x 128(phase), in-plane resolution of 3.6mm x 3.6mm, TR/TE:10000ms/61ms, single spin echo with high order eddy current correction, BW: 250kHz, Slice thickness:6mm, 33 slices/station (i.e. 19.8cm per station), STIR, TI=180ms, b-value=50s/mm² (2 NEX), 800s/mm² (10 NEX), diffusion encoding:3-in-1, scan time: 2min per station. b) The protocol was repeated using a version of the pulse sequence modified to accommodate a 4-shot Single Spin Echo scheme with a 192x192 matrix and higher in-plane resolution of 2.4mm x 2.4mm, partial-Fourier of 9/16, and TR/TE=95500/58ms; b=50s/mm²(NEX=2) and b=800s/mm²(NEX=6), scan time: 5min per station. c) A subsequent dataset was created by discarding shots 2 and 4 of the 4-shot acquisition to simulate an accelerated 2-shot acquisition with large matrix size, referred herein as Fast MUSE, scan time 2min30s per station. All images were collected using the RTB0 distortion correction method to prospectively adjust the center frequency for each slice².

Data reconstruction

All multi-shot images were reconstructed using a MUltishot Sensitivity Encoded algorithm (MUSE)³, which aims at mitigating motion-induced phase errors by solving a linear system P = SD as illustrated in figure 1. P, S, θ, and D respectively correspond to the aliased image from each shot, the sensitivities from each channel, the motion-induced phase maps for each shot and the final unaliased image.

Figure 2 shows the DWI b=50 (top row) and b=800 (bottom row) results for a representative mid-coronal image respectively for the 4-shot MUSE acquisition scheme (left column), Fast MUSE (middle column) and single shot acquisition (right column). First, it is seen that the MUSE algorithm robustly corrects for shot-to-shot motion-induced phase errors across the 4 stations for both the low and high-b value scans. Second, the high-resolution scans exhibit a notable increase of anatomical conspicuity compared to single-shot acquisition as seen in the subcortical, neck, kidney and lumbar spine areas. Third, the Fast-MUSE images have relatively lower SNR compared to the 4-shot acquisition mode, as expected, but retain equivalent levels of anatomical details which are on the other hand superior to the single-shot dwi images.

Our initial results demonstrate that a multishot acquisition scheme based on the MUSE reconstruction algorithm may be a viable option for high-resolution whole-body diffusion imaging. Although the fully sampled high-resolution protocol leads to a two-fold increase in scan time, we showed that multi-station diffusion images can be robustly reconstructed using the Fast MUSE approach, which merely occasions a modest increase in total scan time. Finally, we anticipate that the enhanced conspicuity in both the low and high b-value scans resulting from higher resolution may alleviate the need for an additional T2 sequence, thereby potentially paving the way for a high-throughput imaging workflow.