Introduction

Ultra-low field (ULF) point-of-care MRI systems enable image acquisition in settings previously out-of-reach for conventional scanners. Neonatal clinical care is an important application, as portable systems will allow imaging on Neonatal Intensive Care Units as well as in low to middle income countries with restricted MRI access. However, neonatal brain tissue properties are significantly different from older subjects, and T₁ in particular is strongly field dependent. Whilst T₁ mapping has been performed in neonates at high field [1,2], studies estimating the neonatal brain tissue T₁ at lower fields have been limited to ex vivo tissue samples [3]. In this work we perform in vivo neonatal brain T₁ mapping at 64mT utilising the Hyperfine Swoop® portable MRI system.

Methods

Eight neonates (mean post-menstrual age: 39⁺⁶, 4 clinically referred, 4 healthy controls) were scanned as part of two NHS UK REC approved studies (12/LO/1247, 19/LO/1384). The scanning protocol comprised a pre-scan calibration, localiser, and multiple inversion-recovery 3D turbo spin echo (TSE) acquisitions with the following sequence parameters: TR = 1.5s, TE = 4.62ms, scan time = 3m32s, FOV = 180x220x180mm (RLxAPxFH), res = 2.8mm isotropic, turbo factor (N_TF) = 48, echo spacing (τ) = 4.62ms, BW = 64kHz, TI = 100ms to 600ms in steps of 100ms, with additional data at 700ms and 800ms in 3 subjects.

Data was exported in DICOM format and processed using Matlab (The Mathworks, Natick MA, USA). No image registration was performed prior to fitting. A cost function penalising the squared error between the data and the predicted signal, S(TI), (equation below; M₀ = fully relaxed magnetisation) was minimised using the lsqnonlin routine.
$$S(TI) = M_0\left|1-2e^{-\frac{TI}{T_1}}+e^{-\frac{TR-\left(N_{TF}-0.5\right)\tau}{T_1}}\right|$$
Regions of interest (ROI) measuring 3x3x3 voxels were drawn in the deep grey matter, cortical grey matter, frontal white matter and occipital white matter to obtain per subject average values and capture variations.

Results

Figure 1 shows data from one subject. Fig. 1A shows the four axial slices containing the selected regions of interest for all inversion times, with the centre of the ROIs marked. Figs. 1B and 1C show M₀ and T₁ maps for the same slices. Figs. 1D-G show the raw data and fit for the centre pixel of each selected ROI, with error bars showing the standard deviation of signals in the full ROI. Figure 2 shows the T₁ maps for all subjects ordered by post-menstrual age at scan in weeks. Figure 3 shows the T₁ values plotted against post-menstrual age, showing trends that are consistent with prior relaxometry studies at higher field [2] despite the limited data currently available. Taking the mean ± standard deviations across all subjects, the T₁ values were 510ms ± 114ms (deep grey matter), 587 ± 112ms (cortical grey matter), 949 ± 302ms (frontal white matter), 772 ± 217ms (occipital white matter).

Discussion

The results show that T₁ mapping in neonates is viable at 64mT. Our white matter results differ significantly from the ex vivo results reported by Koenig [3], where T₁ was approximately 494ms at 64mT. In vivo neonatal measurements reveal longer T₁ values than in adults, as observed in local pilot data and at 50mT [4]. T₁ in white matter in some subjects showed large variation, with isolated pixels having very high fitted values. Future work will extend the range of TIs sampled to better measure the longer T₁ values in white matter and will enlarge the cohort to better track age-dependent variation. Additionally, we hope to use these methods to optimise T₁-weighted structural imaging in this population, as well as to identify pathology and measure brain development.

Acknowledgements

We thank Dr Nathan Artz for his input during the initial stages of this project. This work is supported by the Bill and Melinda Gates Foundation and the MRC (Translation Support Award: MR/V036874/1).