In this study we compared the image quality of state-of-the-art 7T brain imaging sequences performed in the same two “traveling heads”¹ between eight different 7T MRI sites.

Two male subjects (37 and 33 yrs.) were imaged at eight UHF sites all operating a 7T whole-body MRI system from the same vendor (Siemens). The systems have differences in basic imaging components, which might influence image quality. Two sites (2, 6) have an actively shielded magnet installed, whereas the rest of the sites use an older, passively shielded magnet. There are also two different gradient coil versions installed at the different sites. The high-performance version (70 mT/m, other version: 38 mT/m) comes with a reduced gradient linearity and is installed at five sites (2, 5-8). At 6/8 sites a commercial RF head coil (Nova Medical) with 1 TX and 32 RX channels was used, while at two sites (1, 4) a similar 24 RX channel version from the same manufacturer was available. To facilitate data analysis, the measurement setup was kept as similar as possible between the sites and the subjects were positioned as standardized as possible with the help of locally available cushions and pillows, as well as auto-align localizer scans.

The imaging protocol consisted of a spin-echo based B1 mapping sequence (TA: 0:27, 8x8x5mm³) for transmitter calibration verified by a DREAM² B1-mapping sequence (TA: 0:05, 5x5x5mm³). Subsequently, the following imaging sequences were performed: MP2RAGE (TA: 9:38min, 0.75×0.75×0.75mm³), TSE (TA: 7:05min, 0.3×0.3×2.0mm³), modified TOF³ (TA: 6:41, 0.2×0.2×0.4mm³), and SWI (TA: 9:05, 0.3×0.3×1.0mm³). An EPI sequence with sinusoidal readout (TA: 4:22, 1.3×1.3×1.5mm³) was used to acquire functional data during 4 minutes scan time with no task.

All measurements were acquired within 1 hour per subject. Data were analyzed qualitatively and quantitatively after co-registration between systems. Registration and image processing was done with FSL (FMRIB Software Library v5.0). MP2RAGE images were registered to the MNI template. The TSE data were registered to MP2RAGE to use VOI masks gained from previous analysis of MP2RAGE. SWI and TOF measurements were co-registered to a second reference scan of each subject taken at Site 3. ROIs were defined in multiple brain structures, and image contrasts were compared between all sites. For anatomic sequences (MP2RAGE and TSE), the contrast, T1, and volume of gray and white matter, CSF and subcortical nuclei were compared. Images from TSE were analyzed in two subgroups for each RF coil, as sequence parameters had to be changed for the 24ch coil. For the angiographic sequences (TOF and SWI), vessel contrast against surrounding background was used for quantification. Temporal SNR (tSNR) maps calculated from EPI as well as B1 and B0 field maps were analyzed for each brain. Additionally, both subjects were rescanned at Sites 2, 3 and 4 to assess intra-system variability.

MP2RAGE images showed very high agreement in contrast and measured T1 values (Fig.1, Table 1). Maximum variation of calculated brain size between the sites was 6% for Subject 1 and 3% for Subject 2. Rescans at 3 sites showed higher precision (1% difference) for same-system data, indicating systematic differences between the sites. T1 values were within 3% variation, except for inferior regions like the brainstem, where the signal level typically drops off with the RF coils used. On TSE images (Fig. 2) one can clearly distinguish the different RF coils. The gray to white matter contrast differences between sites were below 8%. However, in central subcortical regions larger contrast deviations (5–20%) were found between the sites with different RF coils (Table 1). The TOF angiography (Fig.3) analysis showed high agreement except for 5/20 vessels, where measured contrast was more than 10% lower for the 24ch coil. SWI image data (Fig.4) had higher vessel to background contrast deviations when compared to a reference scan, inter-site comparison showed a difference between 5% and 18% depending on vessel position and size.UHF MR systems are very sensitive measurement devices⁴. Even when provided from the same vendor, there are differences between individual devices and sites that potentially might influence image quality. Effects of hardware differences such as RF coils were revealed, but we did not control for physiological differences in the individual subjects such as heart rate or breathing frequency that might also influence image quality. Nevertheless, our results show, that inter-site comparability of UHF measurements can be achievable in the near future. This is a promising finding for the comparability of studies, such as fMRI or relaxometric quantification of disease markers, and does ensure the same diagnostic reliability on all MRI devices.