Development of a real-time fetal brain tracking method to be applied to BOLD MRI and diffusion MRI aimed at monitoring cerebral development in-utero

胎儿大脑MRI提供了对出生前生活的独特见解。胎儿BOLD和扩散MRI测量与胎儿的健康和生长相关，因此提供了一个机会，探测过程中涉及人类发展的非侵入性。胎儿MR的主要挑战之一是不可预测的胎动。通常选择多切片2D采集。因此，快速单次激发采集允许在每个切片/体积的读出期间冻结运动。后处理技术（SVR）允许从这些切片重建连续的3D体积。然而，依赖于在以下分析中组合的多次重复采集的技术，例如BOLD MRI和dMRI，受到体积间运动的显著影响，并且通常需要后处理以改善时空分析。因此，这些MR技术将从用于运动鲁棒性的大脑跟踪中高度受益。这种方法已在成人中实施- 3D-EPI导航仪被采集和处理用于扫描到扫描的前瞻性运动校正。然而，由于周围的子宫环境和母体组织，其扩展到胎儿成像是具有挑战性的。虽然已经示出了胎儿3D-EPI导航器的示例，但是它们易于出现运动伪影，并且它们的处理涉及手动干预和缓慢的常规配准。因此，我们正在开发一种固有的运动鲁棒性胎儿MRI方法，该方法使用3D U-网进行胎儿头部位置和平移运动估计。因此，我们的方法直接使用采集体积作为导航器。这是可能的，因为这种采集的速度相对较高（TR<5 s）。通过Gadgetron的快速实时反馈系统将为扫描仪提供大脑位置估计，以在每次动态之前更新VOI位置并保持它而不管胎儿运动。我们的目标是这种方法来监测子宫内的大脑发育，并在常规的临床评估中采用。

Fetal brain MRI offers unique insights into life before birth. Fetal BOLD and diffusion MRI measurements are correlated with fetal health and growth, thus offering an opportunity to probe processes involved in human development noninvasively. One of the main challenges of fetal MR is unpredictable fetal motion. Typically multi-slice 2D acquisitions are chosen. Thereby, the rapid single-shot acquisitions allow freezing of the motion during the readout of each slice/volume. Post-processing techniques (SVR) allow reconstruction of a continuous 3D volume out of these slices. However, techniques that rely on the acquisition of multiple repetitions all combined in the following analysis, such as BOLD MRI and dMRI, are significantly affected by inter-volume motion and usually require post-processing to improve spatiotemporal analysis. Therefore, these MR techniques would highly benefit from brain tracking for motion robustness. Such methodology has been implemented in adults- 3D-EPI navigators are acquired and processed for scan-to-scan prospective motion correction. Its extension to fetal imaging is, however, challenging due to the surrounding uterine environment and maternal tissue. While examples of fetal 3D-EPI navigators have been shown, they are prone to motion artifacts and their processing involved manual intervention and slow conventional registration. We are therefore developing an intrinsically motion-robust fetal MRI method that uses a 3D U-Net for fetal head position and translational motion estimation. Our method thereby uses the acquisition volumes directly as navigators. This is possible due to the relatively high speed of such acquisitions (TR<5 s). A fast real-time feedback system, via Gadgetron, will provide the scanner with the brain position estimation to update the VOI position before each dynamic and maintain it irrespective of fetal motion. We aim for this method to monitor brain development in-utero and to be adopted in routine clinical evaluation.