As the use of ultra-high field (UHF) MR imaging expands, there is an increasing need to establish high-resolution imaging protocols for clinical use. Magnetization transfer (MT) imaging^[1] has been used to provide information about tissue structure and pathological changes^[2-4]. Because of the much higher specific absorption rate (SAR) of tissue and longer acquisition time, however, in-vivo studies using MT at UHF have not been clinically feasible. We have proposed variable density magnetization transfer(vdMT) technique, and demonstrated the feasibility of MT scan at 7T^[5,6]. In this study, we 1) investigated effects of varying MT densities in the center of k-space and TRs, and 2) demonstrated whole brain MT scans of Multiple Sclerosis (MS) patients using vdMT at 7T in a clinically reasonable scan time.Data were collected from an agar phantom^[7], four controls, an MS postmortem brain and four MS patients in a 7T MRI (Siemens; IRB approved). Figure 1 shows how k-space is acquired in vdMT. To reduce SAR while maintaining similar MT saturation, along the kz-direction the MT pulses are applied every TR in central k-space and sparsely applied in peripheral k-space region. Consequently, vdMT results in lower SAR than conventional MT. To investigate vdMT signal characteristics, two experiments using an agar phantom were conducted. Data were acquired with combinations of TR and MT density as a function of MT offset frequencies (–10kHz~ 0Hz). For vdMT imaging parameters optimization, four sets of vdMT (MT_H50L5, MT_H40L5, MT_H30L5, and MT_H20L5) and a conventional MT data were acquired from each volunteer. Minimum TR (within SAR limitation) was used. To demonstrate the spatial distribution of the MTR map using high-resolution vdMT, data from postmortem MS brain and MS patients were acquired with optimized parameters (MT_H40L5, TR/TE=45/3.5ms, 104 slices, (0.75mm)³-isotropic voxel and GRAPPA factor=3). Parameter-matched GRE data were acquired for MTR calculation. MT saturation (MT_SS=100×[M_S/M₀] where M_S and M₀ represent values with and without MT-prepared signal intensity) and MTR (=[M₀-M_S]/M₀) maps were generated. After generating MTR maps, non-uniform B₁ induced error was corrected^[8]. To measure similarity between conventional method and vdMT, a voxel-wise correlation was calculated.

As the level of MT density in the central k-space increased, the amount of signal saturation in the vdMT increased (Fig. 2A). All vdMT scans with the same TR as the conventional MT provided lower MT saturation than conventional MT. However, MT_H40L5 demonstrated comparable MT-saturation compared with the conventional MT. This suggests that the amount of MT-saturation can be increased with increasing dense MT RF area in the central k-space. When the TR was decreased, the amount of MT-saturation was increased (Fig. 2B). vdMT using the same TR as the conventional method provided the lowest MT-saturation. In contrast, the shortest TR generated the highest MT-saturation. These results show that a shorter TR in vdMT leads to a higher MT-saturation, suggesting that MT-saturation is maintained over several TRs if TR is short.

MTR maps using vdMT from controls demonstrated 18.3% higher tissue contrast than conventional MT. Among the MTR maps which show higher white matter MTR value than conventional MT, MT_H40L5 demonstrated the nearest MTR value to the conventional MT. Hence, the level of MT dense area is optimized with H=40. Results in Fig. 3 point to the similarities between vdMT and conventional method in image quality. When voxel-wise correlation was performed, the mean correlation coefficients were 0.9. These results suggest a high degree of similarity between the two methods. When the MTR map from the postmortem brain was qualitatively investigated, any noticeable artifacts were not seen. Figure 4 shows MT_H40L5 images from T₁-weighted, FLAIR, T₂*-weighted, and MTR maps of a postmortem MS brain (A-H) and of an MS patient (I-P). MS lesions demonstrated hypo signal in the T₁-weighted image and hyper signal in the FLAIR and T₂*-weighted images. The corresponding areas in the MTR maps demonstrated significantly reduced signal levels, clearly delineating lesions.

In this work, we demonstrated a new approach for acquiring whole brain covered 7T MT data in a clinically reasonable time. Our vdMT method provides similar image quality to that obtained with conventional MT imaging, and shortens the scan time, or TR by avoiding from SAR limitation, which minimize the incidence of motion artifacts. The proposed method generates high-resolution MT data in reasonable scan time and it exhibits high similarity with the conventional method. Moreover, it maintains sensitivity to MS lesions. These features make the vdMT method appealing for clinical neuroimaging applications in UHF.