Next-generation in-vivo fetal neuroimaging

下一代活体胎儿神经成像 该项目的总体目标是显著改进胎儿磁共振成像(MRI)，以 人类早期大脑发育和神经发育障碍的研究进展，其负担 不幸的是，由于其终生影响和高流行率，这一比例很高。胎儿核磁共振一直是一项技术 是研究胎儿期大脑发育的首选。然而，胎儿运动使MRI切片采集 充其量是不可靠的，因为在对规定的切片进行成像时，胎儿经常移动。未代偿胎儿 运动破坏了解剖的3D覆盖，并降低了切片与体积的空间分辨率 重建。重复扫描并不能确保解剖结构的完全3D覆盖，但会增加总体 获取时间。这反过来又大大降低了胎儿核磁共振研究的成功率和可靠性。 胎儿一过性脑室发育过程中对损伤的选择性敏感 胎儿发育。为了缓解这些问题并改进胎儿MRI，我们建议自动测量 胎儿大脑位置和预期实时导航切片到每个新位置。这件事的影响 方法是显著提高活体胎儿磁共振成像的成功率和空间分辨率 脑室发育的研究，同时，减少扫描时间，有效地使胎儿 核磁共振检查给母亲带来的负担更轻，更准确，更具成本效益。通过消除手动重新- 调整切片堆叠位置后，两次扫描之间经过的时间实际上将是连续的。我们的 拟议的技术还将使胎儿MRI对操作员的依赖减少，从而在整个过程中更具重复性 网站，这是进行多中心研究和临床试验所必需的。胎儿的前瞻性导航 MRI切片用于补偿运动需要开发新的、实时的图像处理算法 识别胎儿大脑及其位置和方向；跟踪胎儿运动以控制切片；并检测 并重新获取运动损坏的切片。在这个项目中，我们将开发创新的深度学习模式，以 实时处理胎儿MRI切片；将这些模型转换为集成系统，以前瞻性地 导航胎儿核磁共振切片；并将在不同孕龄扫描的胎儿上验证该系统。至 在评估拟议技术的实用性和影响时，我们将测量胎儿的底板体积。四个 本研究的具体目的是：1)通过可变密度图像采集和重建来评估胎儿MRI； 2)在MRI切片上实现了对胎脑的实时识别；3)开发了胎头实时识别系统 脑片的运动跟踪和控制；4)测量发育中胎儿大脑的亚板体积 核磁共振检查。这些目标将共同转化和验证新的成像和图像处理技术，以 推进胎儿核磁共振，有效消除在发育领域取得进展的关键障碍 神经学和神经科学。

Next-Generation In-Vivo Fetal Neuroimaging The overall objective of this project is to dramatically improve fetal magnetic resonance imaging (MRI) to advance research in early human brain development and neurodevelopmental disorders, the burden of which is, unfortunately, high because of their life-long impact and high prevalence. Fetal MRI has been the technique of choice in studying prenatal brain development. Fetal motion, however, makes MRI slice acquisition unreliable at best, as the fetus frequently moves while the prescribed slices are imaged. Uncompensated fetal motion disrupts 3D coverage of the anatomy and reduces the spatial resolution of slice-to-volume reconstructions. Repeating the scans does not ensure full 3D coverage of the anatomy, but increases total acquisition time. This, in turn, dramatically reduces the success rate and reliability of fetal MRI in studying the development of transient fetal brain compartments that are selectively sensitive to injury over the course of fetal development. To mitigate these issues and improve fetal MRI, we propose to automatically measure fetal brain position and prospectively navigate slices to each new position in real-time. The impact of this approach will be to dramatically increase the success rate and spatial resolution of fetal MRI for the in-vivo investigation of developing brain compartments, while, in parallel, reducing scan time, effectively making fetal MRI less burdensome for the mother, more accurate, and cost effective. By eliminating the manual re- adjustment of stack-of-slice positions, the time that elapses between scans will be virtually continuous. Our proposed technique will also make fetal MRI less operator-dependent and thus, more reproducible across sites, which is essential to conducting multi-center studies and clinical trials. Prospective navigation of fetal MRI slices to compensate for motion requires the development of novel, real-time image processing algorithms to recognize the fetal brain and its position and orientation; to track fetal motion to steer slices; and to detect and re-acquire motion corrupted slices. In this project, we will develop innovative deep learning models to process fetal MRI slices in real-time; will translate those models into an integrated system to prospectively navigate fetal MRI slices; and will validate the system on fetuses scanned at various gestational ages. To assess the utility and impact of the proposed technology, we will measure subplate volume in fetuses. The four specific aims of this study are to 1) assess fetal MRI via variable density image acquisition and reconstruction; 2) achieve real-time recognition of the fetal brain in MRI slices; 3) develop a system of real-time fetal head motion tracking and steering of slices; and 4) measure the subplate volume in the developing fetal brain using MRI. These aims will collectively translate and validate new imaging and image processing techniques to advance fetal MRI, and effectively eliminate a critical barrier to making progress in the fields of developmental neurology and neuroscience.