Introduction

Myeloarchitecture has a critical role in the specialisation of brain microcircuitry. It shapes network functions from small to large scale, and ultimately behaviour [1]. State-of-the-art MRI can potentially allow direct observation of the bootstrapped development of patterned myelination in the cortex [1]. Recently, Sanchez and co-workers proposed to identify specific myeloarchitecture by mapping T1, by means of slice-shuffled interleaved multi-shot inversion recovery EPI (MS-IR-EPI) at 7T [2,3]. At sufficient spatial resolution (better than 500μm), this approach has some advantages when compared with other commonly-used T1 mapping methods (e.g. MP2RAGE [4]). Use of EPI for T1 quantification is reliable and is often used as a gold standard [4], it can minimise the effect of T1 relaxation on the point spread function (PSF) [5], and its SNR per acquisition time can exceed that of many other techniques. In the readout direction, PSF is only significantly affected by k-space truncation and discrete sampling. In the phase-encode direction, however, relaxation effects are not negligible. In IR-EPI, transverse relaxation occurring during the echo train can cause some broadening of PSF. This effect can be minimised by using more EPI segments, at the expense of total acquisition time. Effects due to T1 relaxation, occurring only through radiation damping [6], can be neglected. By contrast, when very high spatial resolution is required for MP2RAGE, the k-space sampling in the second phase-encode direction (phase2) takes place during a significant proportion of the inversion recovery curve, and T1 relaxation can significantly affect the PSF. In the current investigation, we explored the reproducibility of MS-IR-EPI T1 maps. Using homemade phantoms, we also estimated and tested the quality of the PSF for both MS-IR-EPI and MP2RAGE T1 maps.

Methods

Phantoms were created by filling 15ml Falcon^TM tubes with different concentrations of agar/agarose gel and CuSO₄ or MnCl₂ (Sigma Aldrich), in deionised water. Acquisitions were performed on a whole-body 7T research MR-system (Siemens Healthcare GmbH, Erlangen, Germany) equipped with a 32-channel head receive/volume transmit coil (Nova Medical). A prototype MS-IR-EPI [3], a Turbo Spin Echo (TSE) and an MP2RAGE were used (see Fig.1). Post-processing computations were performed under Jupyter Lab and MATLAB R2014b on a Mac Pro running OSX 10.11. MS-IR-EPI T1 maps for six phantoms were computed voxelwise by fitting the IR relaxation equation to the data acquired with 16 equally spaced inversion times(see Fig.2a). T1 values were averaged on manually drawn ROIs. Reproducibility was assessed by computing the mean value and standard deviation (StDev) across 8 repetitions. One phantom (#14) was formed by filling the vial with two different gel compositions, separated by a thin transition layer (see Fig.2). A TSE image with 100µm in-plane resolution was acquired on this phantom to depict the boundary profile (Fig.2b). MS-IR-EPI and MP2RAGE profiles across the gel boundary for this phantom were measured on four transects of the boundary near the centre of the phantom. We estimated the theoretically expected image profiles for the phase2-across-gap MP2RAGE and phase-across-gap MS-IR-EPI protocols, by calculating analytically the optical transfer function (OTF) in the presence of transverse and longitudinal relaxation. A discrete Fourier Transform of this OTF, which involves the k-space truncation and sampling effects, was used to determine the PSF. The PSF was convolved with a step function representing the gel boundary in phantom #14, to provide an estimate of the MRI signal profile across the gel boundary

Results

The reproducibility of the T1 map is shown in Fig.2a. Fig.2b shows the gel transition boundary profile of phantom #14 determined by high-resolution TSE. Fig.3 shows the measured T1 map profiles of phantom #14. Fig.4 shows theoretical estimates of the OTF and PSF for data acquired with the phase-encode direction across the boundary. Fig.5 shows the modelled effects of low flip-angle 6/8 partial Fourier pulse trains on the longitudinal relaxation after an inversion pulse. Our simulations clearly show the expected narrow PSF of MS-IR-EPI PSF. The MP2RAGE data shown in Fig.3 confirm the expected improved sharpness for a readout direction across the gel boundary, as compared with 6/8-partial-Fourier-phase direction across the boundary.

Discussion

For very high-resolution T1 mapping (better than 500µm) using sequences with inversion pulses and long trains of low flip-angle acquisitions, T1 relaxation can impair the PSF. In MP2RAGE, for instance, although the longitudinal relaxation is only weakly affected by the low flip-angle RF pulses, (see Fig.5a), the relaxation effect on PSF cannot be ignored (see Fig.5b) and shows up in the poor visibility with MP2RAGE of cortical layers of myelin. Attempts to minimize this effect with partially parallel or partial Fourier acquisition unfavourably reduces the already critical SNR per unit time. By contrast, high-resolution MS-IR-EPI provides an optimally narrow PSF when the acquisition time and echo train length are appropriately set. Importantly, the flexibility of the MS-IR-EPI approach allows margins of improvement, affecting only the acquisition time, that is unavailable for other approaches.

Conclusion

The recently proposed MS-IR-EPI based approach to map myelin patterns in the GM shows good reproducibility when tested on homemade phantoms and appears suitable to improve the mapping of grey matter myelination at the nominal (desired) acquisition resolution, thanks to the preserved quality of its PSF at resolutions better than 500µm.

Acknowledgements

No acknowledgement found.