Towards Higher Spatial Resolution Echo-Planar-Imaging With A Compact Head 3T System
Ek T Tan1, Seung-Kyun Lee1, Paul Weavers2, Matthew Middione3, Matt A Bernstein2, John Huston2, and Thomas KF Foo1

1GE Global Research, Niskayuna, NY, United States, 2Mayo Clinic, Rochester, MN, United States, 3GE Healthcare, Menlo Park, CA, United States


It is challenging to increase spatial resolution (≤1.5 mm) in whole-brain, single-shot echo-planar-imaging (ss-EPI) on conventional whole-body systems due to EPI distortion and limited SNR. A compact head 3T system with an asymmetric head gradient coil capable of high gradient amplitude and 3.5 times the slew rate of whole-body systems can enable ss-EPI acquisition with high spatial resolution and reduced spatial distortion, simultaneously. This work compares spin-echo and gradient-recalled-echo ss-EPI between the compact and whole-body systems, showing substantially reduced distortion and signal dropout, and shorter echo-times. Results with the high performance gradient were also demonstrated in multi-band-accelerated, high b-value diffusion-imaging.


Single-shot echo-planar imaging (ss-EPI) is routinely utilized for diffusion imaging and fMRI due to its fast readout. Improvements to spatial resolution are however limited by SNR and EPI distortion. In-plane parallel imaging is frequently utilized to reduce EPI distortion at the cost of reduced SNR. As compared to conventional whole-body systems, a compact head 3T system1 with a dedicated, high-performance head-gradient system2 has superior gradient slew rate (SR) performance by 3.5 times (Gmax=80 mT/m, SR=700 T/m/s), allowing echo spacing (ESP) to be reduced by up to a factor of two without encountering peripheral nerve stimulation limits3-4. The purpose of this work is to compare spin-echo (SE) and gradient-recalled-echo (GRE) EPI performance between the compact 3T and whole-body scanners, and to evaluate the high performance gradient of the compact scanner in advanced diffusion imaging.


The compact 3T system is built around a conduction-cooled, low-croygen magnet that has a warm-bore inner diameter of 62 cm, a magnet weight less than 2,000 kg, a 5 Gauss fringe field area of 24m2, and an imaging FOV (DSV) of 26 cm. Healthy human subjects (under an IRB-approved protocol) were imaged using the compact 3T system, and on a conventional whole-body 3T MRI system (GE MR750). 2D SE-EPI and GRE-EPI axial acquisitions were performed at 1.5-3.4 mm-isotropic sampling (FOV=20.8 cm, TR/TE=1500/40 ms) without in-plane parallel imaging. The ss-EPI readout amplitude was set to 51mT/m on the compact 3T, which allowed full-sampling of the phase-encoding k-space while keeping under TE=40 ms in GRE-EPI. GRE-EPI with the whole-body 3T was performed with partial-ky=0.75, as were all SE-EPI scans on both systems.

Whole-brain diffusion-imaging was also tested, using the high performance head-gradient inserted in a whole-body 3T magnet (GE MR750w), to evaluate the feasibility of acquisition at 1.5 mm-isotropic spatial resolution. Simultaneous multi-slice acquisition5 with no in-plane parallel acceleration (R=3×1) and with slice-select gradient reversal6 was used to shorten acquisition time without residual fat artifacts from the fat saturation pulses. A 32-channel brain coil (Nova Medical) was used in all experiments. The acquisitions tested were: compressed-sensing diffusion spectrum imaging (CS-DSI)7 (bmax=8000 s/mm2, TR/TE=3800/75.1 ms, CS-acceleration R=4, 6 T2 + 127 acquired gradient directions, scan time=8.5 minutes) and multi-shell HARDI8 (bmax=3000 s/mm2, TR/TE=3800/71.4 ms, two b-values with total of 6 T2 + 98 acquired gradient directions, 6.6 minutes). No EPI-distortion correction was applied in the image reconstruction. The readout gradient was reduced to 40 mT/m to increase image SNR, at the expense of slightly increased ESP=456 µs (vs 388 µs at 51mT/m).


For all spatial resolutions tested (1.5-3.4 mm), the ESP with the compact 3T EPI scans was approximately half that with the whole-body scanner (Table 1). The ESP for 1.5 mm EPI in the compact scanner was shorter than that of 3.4 mm EPI with the whole-body scanner. At 1.5 mm, there was visibly reduced susceptibility-induced distortion and signal dropout in both SE-EPI and GRE-EPI with the compact scanner (Figure 1), especially in the vicinity of air-tissue-bone interfaces. However, there was increased residual fat ghosting. The compact scanner produced images with low distortion at all spatial resolutions (Figure 2). Moreover, the extent of signal dropout in the GRE-EPI due to dephasing was reduced with thinner slices.

Diffusion imaging at 1.5 mm-isotropic with multi-shell and DSI provided adequate SNR, sufficient for visualizing colorized FA maps and diffusion kurtosis9 (Figure 3), as well as the major fiber bundles as processed using orientation distribution functions that resolve fiber-crossings (Figure 4).

Discussion and Conclusion

Preliminary imaging experience using 2D ssEPI on the compact 3T system equipped with the high performance head-only gradient suggests the feasibility of routine acquisition of 1.5 mm in-plane resolution with both spin-echo and gradient-recalled-echo. In particular for GRE, the short echo spacing allows for full-echo sampling at TE≤40 ms with low signal dropout, which makes high-resolution fMRI feasible at 1.5 mm. The combined advantages of reduced ESP and TE in diffusion imaging may enable high-resolution and high b-value acquisitions. Susceptibility artifacts are reduced primarily by the decreased ESP, and secondarily by the smaller voxel volumes, which are better supported by the increased gradient performance of the compact system. The EPI ghosting may be reduced with further optimizations of the gradient driver. The low image distortion provided by the compact 3T system provides future avenues for researching rapid ssEPI-based sequences to replace conventional workhorse pulse sequences such as T1-FLAIR and T2-FLAIR.


This work was supported in part by NIH R01EB010065. The views herein do not necessarily represent those of NIH. The authors thank John Andrew Derbyshire for useful discussion.


1. Foo TKF, et al. ISMRM 2016.

2. Huston J, et al. ISMRM 2015. 971.

3. Lee SK, et al. Magn Reson Med (Accepted).

4. Lee SK, et al. ISMRM 2014. 310.

5. Setsompop K, et al. Magn Reson Med 2011. 67(5):1210-24.

6. Middione M, et al. ISMRM 2015. 958.

7. Menzel MI, et al. Magn Reson Med 2011. 66(5):1226-33.

8. Aganj I, et al. Magn Reson Med 2010. 64(2):554-66.

9. Jensen JH, et al. Magn Reson Med 2005. 53(6):1432-40.


Table 1: Summary of pulse sequence parameters for spin-echo and gradient-recalled-echo EPI at various spatial resolutions for an imaging field-of-view of 20.8 cm.

Figure 1: GRE-EPI and SE-EPI at 1.5mm resolution. EPI distortion and signal drop out were greater in the whole-body 3T images, as compared to the compact 3T images (red arrows). However, there was increased residual fat ghosting in the compact 3T images (yellow flat-arrows).

Figure 2: GRE-EPI and SE-EPI compact 3T scans at one axial slice location at various spatial resolutions. Low spatial distortion was observed in all images. Increased signal dropout was seen in the GRE-EPI images at increased slice thickness due to signal-dephasing.

Figure 3: Sagittal and coronal reformats of 1.5 mm-isotropic, 2D axial, multi-shell diffusion imaging (6.6 minutes, 6 T2 + 98 gradient directions) with R=3 simultaneous-multi-slice acceleration, acquired on the head-only gradient (80mT/m, 700T/m/s), showing averaged b={0,1500,3000}s/mm2 maps, colorized FA maps, and orthogonal kurtosis maps, processed without EPI distortion correction.

Figure 4: Diffusion tractography of CS-DSI at 1.5 mm-isotropic resolution (8.5 minutes, 6 T2 + 127 gradient directions), acquired with the head-only gradient (80mT/m, 700T/m/s) and with R=3 simultaneous-multi-slice acceleration, showing renderings of (a) axial-cephalad view, (b) callosal fibers, and (c) major fiber bundles, as visualized with Trackvis (MGH).

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)