T1 relaxometry and tissue segmentation of the human brain at 9.4T and 3T using MP2RAGE

Purpose/Introduction: Accurate and precise determination of T1 values is of central importance in clinical studies and for tissue segmentation based on the myeloarchitecture that transcends T1 [1]. The MP2RAGE sequence is attractive for use at high field since the resulting T1 values are valid across a large range of transmit field inhomogeneities [2]. Here we investigate the use of this technique for whole brain segmentation and T1 relaxometry at 9.4T and compared the results to data obtained 3T. Subjects and Methods: The study was ERB approved and MRI was performed in healthy volunteers (21–56y N = 11) at 9.4T using a 16ch transmit/31ch receive array [3], at 3T a 2ch body transmit coil 32ch receive array. At 9.4T, B1 + mapping with the Actual Flip Angle method [4] and MP2RAGE (TI1/TI2 = 900/3500 ms; FA = 4/6; TR_GRE = 6 ms; TR = 8894 ms, 0.8 mm isotropic voxel size) were performed. The sequence parameters used at 3T were slightly different (TI1/TI2 = 700/2500 ms; FA = 4/5; TR_GRE = 7.7 ms; TR = 5000 ms, 0.8 mm isotropic voxel size). Standard m ethods for automatic brain segmentation in SPM12 based on T1 weighted images or flip angle corrected T1 maps and the IBA SPM tool were employed to define brain areas for analysis. Results: The actual flip angle observed in vivo deviated by 40 and more from the nominal angle so the intrinsic bias correction of the MP2RAGE method failed in brain regions with the highest B1 field. As a consequence, the outcome of brain segmentation deteriorated, leading to mis-classifiation of several grey matter voxels, especially those located in sub cortical areas. Flip angle correction of the MP2RAGE maps acquired at 9.4T substantially improved segmentation and yielded grey matter classification highly similar to the one at 3T (Fig. 1). T1 values extracted in IBA segmented areas showed high consistency across subjects at both field strengths with standard deviations below 3.5 , but showed a greater dynamic range at 9.4T. Discussion/Conclusion: The intrinsic bias correction of MP2RAGE does not suffice to achieve reliable T1 quantification at UHF. Provided transmit field correction is performed, T1 maps with high intersubject consistency is obtained at 9.4T, holding promise for investigation of local myeloarchitectonics in neuroscientific and clinical studies