Susceptibility-weighted imaging (SWI) is sensitive to susceptibility changes associated with blood and and shows excellent performance in detecting haemorrhage. Preterm infants and those with congenital heart disease are at risk of parenchymal, cerebellar or intraventricular/germinal layer haemorrhage. As such, there is a clinical need for rapid, high resolution SWI to assess brain injury in these infants.

Some literature has shown its application into neonatal imaging[1,2], however, to the authors knowledge, no dedicated infant optimization has been presented to the scientific community. Therefore, we sought a dedicated neonatal SWI protocol that maximizes the obtainable signal while minimizing the expected WM/GM contrast and improve sensibility to local field variations within a 5min time constraint.

Relaxation times in the neonatal brain differ from those found in adult brains. In a spoiled gradient echo (SPGR) sequence for a given T1/T2 (tissue-of-interest) the signal is mainly dependent on repetition time (TR), echo time (TE) and applied flip angle (FA) following the well-known Ernst relationship:

$$sin(FA)\frac{1-e^{-\frac{T_R}{T_1}}}{1-e^{-\frac{T_R}{T_1}}cos(FA)}e^{-\frac{T_E}{T_2}}$$

The following design constraints were adopted:

1. Maximize WM/GM signal;

2. Minimize WM/GM contrast;

3. Maximize TE to enhance susceptibility contrast;

4. Minimize the CSF signal;

5. Keep total acquisition time within the 5min limit;

We fixed our acquisition to a fully flow compensated 3D SPGR sequence with a defined FOV (AP,FH,RL) of 180mmx90mmx180mm, an acquired matrix of 360x45x360 and a SENSE factor of 2. Given the self-imposed time limit of 5min, this matrix implies that the allowed TR is less than 32ms. FA variations were explored between 1° and 50°. Maximum echo time, required to maximise susceptibility weight was enforced by fixing TE=TR-7ms to allow time for RF and necessary gradients.

The expected signal variation for CSF, WM and GM were then plotted alongside the WM/GM contrast prediction as a function of TR and FA in order to identify the optimal operation point. Reference relaxation times obtained from[3] are CSFT1/T2=4100ms/2000ms, WMT1/T2=2500ms/240ms and GMT1/T2=1800ms/144ms. Simulation results can be found in Figure1.

With the design objectives in mind a test-range (black line) of TR between 28ms (shorter scan, lower CSF signal, higher GM-WM contrast) and 32ms (longer scan, higher CSF signal, lower GM-WM contrast) was considered. A prescribed FA of 12⁰ was selected as this balances reduced CSF signal (systematically allows us to avoid its Ernst peak) which drives to larger flip angle with keeping WM-GM contrast low, which drives towards lower flip angle,while maintaining the WM and GM close to maximum signal.

3D SPGR acquisitions using this range of parameters were obtained on a Phillips 3T Achieva Tx scanner in 12 new-born babies, all less than one week old. Due to the difficulties of neonatal imaging, such as subject motion, the different settings could not be acquired in a single subject. Informed parental consent was given for all subjects according to local ethical approval requirements.

After acquisition, data was exported and post-processed offline. Phase images are high-pass-filtered to produce a positive phase-mask as discussed in[2] to enhance susceptibility weighting of the magnitude images. Minimum-intensity-projections (mIP) were also produced. The images were jointly assessed with the local clinical team to select which set of parameters would produce most clinically relevant images.

The optimal acquisition protocol was considered as TR=32ms and FA of 12° which lies within the range of optimal acquisition parameters for adult acquisition suggested in[3]. From Figure1 we expect this to be the highest SNR image, with the disadvantage that it results in the longest acquisition time, although still within the 5min constraint. It was qualitatively found that loss in SNR with other parameter choices was not worth the gain in acquisition time. For reference exemplar magnitude, phase, SWI and mIP are shown in Figure2.

Careful inspection of Figure2 shows, as expected, minimal WM/GM contrast in the magnitude image enhancing therefore the good contrast between tissue and venous vessels.

Clinical applicability to assess brain heamorage is compared to clinical standard T2w scans and can be seen in Figure3 where the increase sensitivity is shown.

In this work a dedicated neonatal SWI protocol was sought by the Ernst relationship for the expected relaxation times in the neonatal brain. A TR=32ms and FA=12° produce minimal WM/GM contrast within an acquisition time of 4min:9seconds. This however lies within the range of optimal acquisition parameters proposed in [3] showing that a single unified protocol can be applied to both adults and neonates.

This protocol is currently being use in a larger study in order to assess brain injury in newborns with congenital heart disease where over 30 babies have been assessed.