Joelle E Sarlls1 and S. Lalith Talagala1
1NINDS/NMRF, National Institutes of Health, Bethesda, MD, United States
Synopsis
It has been shown that the temporal
SNR (tSNR) of GRAPPA EPI can be improved by using different autocalibration
scan (ACS) acquisitions. We evaluated the impact of using FLASH-ACS for high
resolution, GRAPPA accelerated EPI at 7T.
We compared the tSNR, ghost levels and distortions characteristics of
EPI data reconstructed using SSEPI, MSEPI and FLASH based ACS at different acceleration
factors and resolutions. Results showed that the tSNR of GRAPPA accelerated EPI
improved by 60-100% when using FLASH-ACS data during image reconstruction. FLASH-ACS reconstructions also had less
residual EPI ghost and identical image distortions compared to SSEPI-ACS and
MSEPI-ACS.
Introduction
Current EPI based
fMRI protocols frequently incorporate accelerated parallel acquisition techniques
(PAT) such as GRAPPA and SENSE. These techniques help to reduce EPI distortions
and to increase the number of slices per TR. Recently, it has been shown that
the temporal SNR (tSNR) of GRAPPA EPI can be improved by using a PAT autocalibration
scan (ACS) based on FLASH (1) or FLEET (2) acquisition schemes. In this work, we
evaluate the impact of using a FLASH-ACS for high resolution, GRAPPA
accelerated EPI at 7T. We compared the tSNR, ghost levels and distortions
characteristics of EPI data reconstructed using single-shot EPI (SSEPI),
multi-shot EPI (MSEPI) and FLASH based ACS at different acceleration factors
and resolutions. Methods
All studies were conducted under
an approved IRB protocol using a 7T MRI scanner. Axial brain images of heathy
volunteer subjects (n = 5) were acquired
using a 32-channel receiver array and a head transmit coil with following
parameters: 2D gradient echo EPI (Siemens WIP676b, VB17 software), TR = 2.5 s,
TE = 28 ms, flip angle = 70 degrees, field-of-view = 19.2 cm, matrix = 128x128
or 80x80 (resolution 1.5 mm and 2.4 mm, respectively), slice thickness = 3 mm, 25
slices, and 100 measurements. Different image series were acquired at each
in-plane resolution with GRAPPA acceleration factor R = 2, 3 and 4 with SSEPI, MSEPI
and FLASH-ACS. Note, SSEPI-ACS was not
available for R=4. All images were
reconstructed on the scanner using software provided by the manufacturer using
the local phase correction algorithm for EPI ghost reduction.
Image series were motion
corrected and low order signal drifts were removed from each pixel time series
prior to calculating the tSNR. The whole brain average tSNR within a brain mask
was computed for each series. The residual EPI ghost levels in manually drawn
ROIs, outside the brain and along the phase encoding direction on 4 evenly
spaced slices, were quantified as percentage of the mean brain signal. Possible
image distortions arising from differences in echo-spacing of each ACS method
were examined by generating brain contours using an edge-detection algorithm.
Results
Figures 1 and 2 show representative tSNR maps for 1.5 mm in-plane
resolution data from the same subject. Images reconstructed with FLASH-ACS have
higher and more uniform tSNR throughout the brain compared to those obtained
using SSEPI-ACS and MSEPI-ACS. SSEPI and
MSEPI-ACS reconstructions exhibit heterogeneous tSNR within and across slices (Fig.
1&2 arrows and sagittal reformats). Figure
3A shows the comparison of the tSNR averaged across the whole brain, for all
subjects. FLASH-ACS reconstructions are seen to produce significantly higher
tSNR for all PAT factors and resolutions compared to SSEPI-ACS and MSEPI-ACS,
with improvement from 60-100% over MSEPI-ACS reconstructions. Figure 4 shows graphs comparing the residual ghost
level, averaged across all subjects. FLASH-ACS
reconstructions have lower ghost level than SSEPI-ACS and MSEPI-ACS for all PAT
factors and resolutions. Figure 5 shows example
images with the most severe distortions, from three subjects, along with
contour lines at the edge of the brain. Data show that edge contours are
essentially identical for MSEPI-ACS and FLASH-ACS reconstructions in all cases. Discussion
In this study, we
have demonstrated that the tSNR of GRAPPA accelerated EPI acquired at 7T can be
improved significantly by using FLASH-ACS data during image reconstruction. An
improvement in tSNR was seen for all PAT factors (R=2, 3, 4) and resolutions
(1.5 mm and 2.4 mm). Reduction of tSNR
when using SSEPI-ACS is attributed to phase errors in the ACS data that
propagate into the GRAPPA kernel calculation (1). Because of the absence of
phase errors, FLASH-ACS data provides a more accurate GRAPPA kernel and thus improved
tSNR in reconstructed images. In
addition, it was seen that the FLASH-ACS reconstructions have less residual EPI
ghost and essentially identical image distortion characteristics compared to SSEPI-ACS
and MSEPI-ACS. Therefore, use of the FLASH-ACS approach should directly benefit
7T fMRI studies using similar protocols, without adverse effects on image
quality. Acknowledgements
No acknowledgement found.References
1)
Talagala et al., MRM 75:2362 (2016)
2) Polimeni et al., MRM 75:665 (2016)