Measuring cortical plate and subplate thickness in the human fetal brain from magnetic resonance images

项目摘要/摘要 在胎儿大脑中，皮质板(CP)的厚度被认为与细胞的数量和大小有关。 柱状、堆积密度、皮质内髓鞘和突触，以及与 丘脑和皮质传入神经的数量以及皮质-皮质连接的数量。大脑皮层的估计 出生后MRI测量的厚度对我们了解人脑发育和 认知功能与各种脑部疾病的发生和发展。然而，我们的知识和 由于缺乏可用的技术，对人类宫内CP和SP厚度的研究仍然有限 从胎脑MRI自动测量局部CP和SP厚度。与儿童或成人的大脑相比， 胎儿的大脑要小得多，图像对比度也不同。胎脑核磁共振显示效果较差 分辨率，并患有头部运动，导致伪影。因此，准确地提取CP是一项具有挑战性的工作 和SP区域，并在CP和SP表面之间定义几何上合适的厚度。本研究 将开发一种全自动管道，使用多点胎儿MRI提取局部CP和SP厚度 数据集。我们将开发基于脑沟识别的CP和SP分割方法 基于深度卷积神经网络的脑脊液区域识别。在准确分割的基础上，一种 将开发专门针对胎儿大脑的变形模型方法来提取CP 和SP曲面。CP和SP厚度将基于所有 CP和SP曲面。我们将使用不同的成像子集执行可靠性和灵敏度测试 通过移动CP和SP边界创建的相同主题和人造数据。然后，我们将定义增长 18~37孕期典型发育(TD)胎儿各皮质区CP和SP厚度比率 周(GW)。我们假设CP和SP厚度的增长率、最大SP厚度和/或 TD胎儿不同皮质区CP厚度的最大生长GW是不同的。这个 脑部异常(多小脑回和体部发育不全)胎儿CP和SP厚度的生长 在统计上将与TD胎儿的生长情况进行比较。皮质发育畸形和 皮质-皮质连接可能导致多小脑回胎儿CP和SP厚度的变化 胼胝体发育不全。这项研究将为一种新的生物标记物奠定基础，这种生物标记物可以导致 对子宫脑发育正常和改变的机制有更深入的了解。我们的方法发展了 来自拟议的研究将使用基于网络的神经图像计算平台公开分发， 这将使胎儿CP和SP厚度分析有更多的临床应用。

PROJECT SUMMARY/ABSTRACT In the fetal brain, cortical plate (CP) thickness is thought to be related to the number and size of cells within a column, packing density, intracortical myelin, and synapses, and subplate (SP) thickness associated with the number of thalamic and cortical afferents and the amount of cortico-cortical connections. Estimation of cortical thickness postnatally with MRI has contributed greatly to our understanding of human brain development and cognitive function and disease onset and progression in various brain disorders. However, our knowledge and research of human in utero CP and SP thickness remains limited due to the lack of available techniques that automatically measure regional CP and SP thickness from fetal brain MRI. Compared to child or adult brains, fetal brains are much smaller in size and have different image contrast. Fetal brain MRI shows lower effective resolution and suffers from head motion which causes artifacts. Thus, it is challenging to extract accurate CP and SP regions and define geometrically appropriate thickness between the CP and SP surfaces. This study will develop a fully automatic pipeline to extract regional CP and SP thickness using multi-site fetal MRI datasets. We will develop the method for CP and SP segmentation with the identification of sulcal cerebrospinal fluid regions using deep convolutional neural networks. Based on the accurate segmentation, a deformable model method that is optimized and specialized for fetal brains will be developed to extract the CP and SP surfaces. CP and SP thickness will be measured based on vertex-wise correspondence between all CP and SP surfaces. We will perform reliability and sensitivity tests using different imaging subsets within the same subject and artificial data created by moving the CP and SP boundary. We will then define the growth rate of CP and SP thickness in all cortical regions in typically developing (TD) fetuses from 18 to 37 gestational weeks (GW). We hypothesize that the growth rate of CP and SP thickness, the maximum SP thickness, and/or the maximum growth GW of CP thickness will be variable across different cortical areas in TD fetuses. The growth of CP and SP thickness in fetuses with cerebral abnormalities (polymicrogyria and agenesis of corpus callosum) will be statistically compared to the growth of TD fetuses. Malformations of cortical development and cortico-cortical connections may result in altered growth of CP and SP thickness in fetuses with polymicrogyria and agenesis of corpus callosum. This study will lay the foundation for a novel biomarker that can lead to greater insight into the mechanisms of normal and altered in utero brain development. Our methods developed from the proposed study will be publicly distributed using a web-based neuroimage computation platform, which will enable more clinical applications of fetal CP and SP thickness analysis.