Fetal brain MRI suffers from unpredictable fetal motion, limiting the types of MR acquisition schemes practical for use in clinical settings to fast single-shot encoding techniques, such as Half-Fourier-Acquisition Single-Shot Turbo-Spin-Echo (HASTE). EPI-based volumetric navigators (EPI-vNavs) have been embedded into a T₁-weighted MPRAGE acquisition scheme to successfully detect and prospectively correct for head motion each TR ¹, but only postnatally. Standalone EPI-vNavs were successfully used to estimate fetal head motion ², and here we extend prior work by implementing a novel combination of EPI-vNavs and HASTE readouts, thus enabling detection and estimation of fetal head motion in real time (each TR). We demonstrate quantitative measurements of fetal head displacement and rotation per TR, and importantly, that the EPI-vNavs in the HASTE sequence do not compromise HASTE image contrast in the fetal brain.

All fetal scans were performed on a 3T Skyra scanner (Siemens Healthcare, Erlangen, Germany) using spine and body flex receive arrays (30-36 coil elements). Pregnant women signed informed consent forms approved by BCH’s IRB. EPI-vNavs were embedded in the standard HASTE acquisition (Figure 1) where the EPI-vNav volume orientation could be prescribed independently from the HASTE slices. The EPI-vNav was based on a 3D GRE-EPI readout (TR=41ms, TE=13ms, voxel size=5x5x5mm³, FA=3⁰, FOV=30x30x12cm³, fat saturation=ON, TAC=738ms) and it followed the HASTE readout (TE=119ms, voxel size=1.2x1.2x3mm³, FOV=30x30cm², partial Fourier=5/8, GRAPPA R=2). The minimum TR of the combined sequence was 1.5s including a ~350ms gap after the EPI-vNav readout to allow for full M_z relaxation before the HASTE readout (Figure 1). Nevertheless, all scans used TR=2s to safely comply with SAR limits.

Motion detection and estimation from the EPI-vNav volumes followed the procedures described in ². An ROI was identified in the first EPI-vNav volume that included the high-contrast areas of the skull base and cervical vertebral bodies (Figure 4). This ROI was registered to each subsequent volume using the FSL FLIRT tool to measure time series of 3 translational and 3 rotational coordinates. When EPI-vNav volumes were acquired in a different orientation from the HASTE slice slab, dark bands due to spin-history effects were observed, and were manually masked out prior to the registration process.

Figure 2 shows comparisons of a representative slice from the conventional HASTE (Figure 2a) and vNav-HASTE (Figure 2b) acquisitions. Aside from the saturated lipid signals in the vNav-HASTE acquisition (due to fat-saturation in EPI-vNav readouts for improved image quality), the signal intensities and contrast in the fetal brain and surrounding maternal/fetal organs looks equivalent. Figure 3 shows EPI-vNav volumes from three different TRs of the vNav-HASTE acquisition. One can clearly see the dark bands through the EPI volumes, as the HASTE readouts loop through different slices of the fetal brain. These dark bands are due to spin-history effects from the HASTE acquisition, and do not affect the robustness of the registration process or fetal head motion estimation. Figure 4 shows the head translation and rotation plotted against time for one representative vNav-HASTE acquisition. Position was measured at every TR even in the presence of spin history artifacts from the HASTE readouts (Figure 4, yellow arrow). A novel MRI sequence with volumetric EPI navigators in HASTE readouts enables reliable detection and estimation of fetal head motion in each TR. The navigators’ excitation pulses do not visibly impair image contrast of the relevant anatomy in HASTE. In our current implementation, the motion estimates are calculated offline using MATLAB scripts that require manual masking of an EPI-vNav volume to segment the fetal head from the rest of the pregnant abdomen, and processing time for calculation of the 6 motion parameters for one EPI-vNav volume takes ~6mins on an Intel Xeon E50269 processor. Future work will address these limitations with the goal of fetal head motion estimates that are calculated in real time, enabling the possibilities for prospective motion mitigation in HASTE imaging. In addition, future work will also integrate the vNav motion estimates with post-processing algorithms that use multiple HASTE acquisitions of the fetal brain prescribed in at least three different orientations, and apply super-resolution reconstruction techniques to produce a single high-resolution volume of the fetal brain.