A 3D multimodal micron-scale human brain atlas bridging single cell data, neuropathology and neuroradiology

数字化参考大脑，也称为通用坐标框架（CCF）， 与叠加的图谱注释，对神经科学至关重要。承担着 与神经科学的关系与参考基因组和基因组注释与细胞和 分子生物学然而，引人注目的是，人类的这种参考大脑远远落后于 用于非人类模式生物如实验室小鼠的相应CCF。现有 数据集要么具有间隔相对较远的部分，要么缺乏平面内分辨率 微米尺度至关重要的是，现有的数据集与医学的主要领域没有很好的联系 神经放射学和神经病理学。 我们将通过创建前所未有的微米级3D地图集来满足这一需求， 多种MRI模式以及连续连续切片组织学。特别是， 图谱将由尼氏染色、髓磷脂染色和H&E染色组成，具有20微米连续连续切片，以及 大约8000个切片/脑。我们将使用磁带传输方法， 即使在大脑存在断开的碎片的情况下， 分段，并允许将这些部分3D重新组装成3D体积。我们将利用 共配准的MRI和组织学数据，并将创建一个 人脑CCF，其中单细胞转录组和表观基因组数据可以按顺序固定 to create创建a Human人Brain脑Cell细胞atlas图集. 我们将使用机器学习方法来分割这些图像中的细胞和过程， 以算法检测细胞结构边界;这样的机器学习方法还将 用于从MRI数据预测组织学和细胞结构，我们收集的数据作为 训练集我们将把我们的数据免费提供给科学家和医疗专业人员 通过一个在线数据门户，使用多分辨率查看器进行缩放和平移 兆像素图像数据，并将数据存款共享数据存储库中， 其他研究人员也可以使用。我们将把我们的数据与一个独特的在线神经病理学 资源包含超过PB的神经病理学图像，包括H&E染色切片 从冠状面。我们希望我们产生的参考大脑数据将成为 这是人脑高分辨率参考图谱的事实标准。

Digitized reference brains, also referred to as Common Coordinate Frameworks (CCFs), together with superposed atlas annotations, are of central importance to neuroscience. They bear the same relation to neuroscience as do reference genomes and genome annotations to cellular and molecular biology. Strikingly, however, such reference brains for humans lag far behind the corresponding CCFs for non-human model organisms such as the laboratory mouse. Existing data sets either have sections spaced relatively far apart or lack in-plane resolution down to the micron scale. Crucially, existing data sets are not well connected to the major areas in medicine that deal with the human brain, namely neuroradiology and neuropathology. We will meet this need by creating an unprecedented micron-scale, 3D atlas that combines multiple MRI modalities as well as continuous serial section histology. In particular, the reference atlas will consist of Nissl, Myelin and H&E stains, with 20 micron contiguous serial sections, and approximately ~8000 sections/brain. We will do so using the tape-transfer method, which preserves tissue geometry even in the presence of disconnected pieces to the brain being sectioned, and permits 3D reassembly of the sections into a 3D volume. We will utilize diffeomorphic mapping methods to co-register the MRI and histological data, and will create a human brain CCF in which single-cell transcriptomic and epigenomic data can be pinned in order to create a Human Brain Cell atlas. We will use machine learning approaches to segment cells and processes in these images and to algorithmically detect cytoarchitectonic boundaries; such machine learning methods will also be used to predict histology and cytoarchitecture from MRI data, with our collected data as a training set. We will make our data freely available to scientists as well as medical professionals through an online data portal with a multi-resolution viewer for zooming and panning through terapixel image data, and also deposit the data in a shared data repository to make it easily accessible to other researchers. We will connect our data to a unique on-line neuropathology resource containing over a petabyte of neuropathological images, including H&E stained sections from the coronal plane. We expect that the reference brain data we produce will become the de- facto standard for a high-resolution reference atlas for the human brain.