The longitudinal relaxation time (T₁) is related to macromolecular concentration, water binding and water content [1], and is therefore important for tissue characterisation and assessment of pathology. Furthermore, the accurate knowledge of T₁ serves as the basis for several other quantitative MR methods, including in vivo spectroscopy, perfusion imaging and quantitative magnetization transfer imaging.

In the spinal cord, conventional protocols are hampered by low resolution, limited coverage and long scan times, and therefore limited in clinical practise. Currently, the only a study reporting T₁ values in vivo at 3T is limited to a single slice at the C2/C3 cord level [2] and requires ~20 minutes of scan time.

We propose a T₁ mapping protocol for the spinal cord that addresses spatial coverage and scan time limitations by combining reduced field-of-view (FOV) acquisition [3] with a slice-shuffling Inversion Recovery (IR) approach [4]. We show that our protocol achieves robust T₁ mapping of the whole cervical spinal cord in under 9 minutes.

Sequence description:

Zonally Magnified Oblique Multi-slice Echo Planar Imaging (ZOOM-EPI) enables the acquisition of a small FOV without aliasing artefacts of surrounding tissue. However, due to the oblique excitation, a TR>>T₁ is required between contiguous slice excitations for the longitudinal magnetisation to recover. Multi-slice acquisition therefore requires adjacent slices to be split into multiple packages with a long TR between each of the packages (figure 1a). The intrinsic constraint TR>>T₁ can be exploited to perform an IR experiment for T₁ measurement within a clinically feasible scan time, i.e. the base sequence is prepended by an inversion pulse followed by different Inversion Times (TI).

We use a non-selective inversion pulse (i.e. the same pulse is experienced by all slices within a package) to acquire data at multiple TI values in a time-efficient way by shuffling the slice acquisition order within the package and varying initial delay following the inversion pulse over different sequence repetitions (figure 1b).

The interplay between the slice shuffling mechanism and short ZOOM-EPI readout times (~40ms) improves scan time efficiency immensely, in particular for large contiguous slice coverage.

Data Acquisition:

To image 15 cm of the human cervical spinal cord, we split a stack of 30 5mm-thick slices into 5 packages, with 12 different TIs. Sequence details: FOV 64x48mm², 1x1mm² in-plane resolution, reconstructed to 0.5x0.5mm², recovery time T_rec=6s, TIs=50, 250, 450, 650, 850, 1050, 1250, 1450, 1650, 1850, 2050, 2250ms. Scan time was 8min16sec.

4 healthy subjects (3M, 25-28 years) were scanned using the IR-ZOOM-EPI sequence and 8 repetitions of an identical spin echo ZOOM-EPI to estimate the noise standard deviation (SD).

Data Analysis:

All data were registered slice-wise to the first TI volume with a 3 degrees-of-freedom model using FLIRT. A mono-exponential model was fitted to magnitude data using maximum likelihood estimation under the assumption of Rician noise. Mean T₁ values for Grey (GM) and White Matter (WM) were extracted from regions of interest (ROIs) manually drawn on the first TI volume and applied to voxel-wise parametric maps.

Figure 2a shows the 12 images acquired at different TIs in a representative slice. In figure 2b, mean WM and GM ROI signal is plotted against predicted signal when voxel-wise parameter estimates are averaged within ROIs. Figure 3 shows T1 maps at different levels of the cervical spinal cord for a single subject.

Mean T₁ (±SD) for both WM and GM in all subjects are reported in figure 4. The mean (± SD) T₁ for WM and GM was found to be respectively 1096 (±26) ms and 1153 (±24) ms, giving coefficients-of-variation (COVs) of 2.36% and 2.09%.

We have proposed a new scan time efficient T₁ mapping protocol for large coverage in the spinal cord. We demonstrate feasibility of T₁ relaxation mapping over the whole cervical cord in 9 minutes using the standard IR approach with excellent inter-subject reproducibility.

Compared to previous findings in the spinal cord at 3T [2], we found higher values for both WM and GM, closer to the ranges usually reported for the brain [5], and reduced differentiation between the tissue types. However, differences in the acquisition protocols used are known to have an impact on T₁ estimation [6].

Due to the short acquisition time it could easily be added to a routine protocol for spinal cord T₁ quantification. In future work, we will investigate whether the acquisition time can be further decreased, e.g. by reducing the number of TIs for T₁ estimation.