Ex vivo MRI of the brain provides remarkable advantages over in vivo MRI for visualizing and characterizing detailed neuroanatomy, and helps to link microscale histology studies with morphometric measurements. However, automated segmentation methods for brain mapping in ex vivo MRI are not well developed, primarily due to limited availability of labeled datasets, and heterogeneity in scanner hardware and acquisition protocols. In this work, we present a high resolution dataset of 37 ex vivo post-mortem human brain tissue specimens scanned on a 7T whole-body MRI scanner. We developed a deep learning pipeline to segment the cortical mantle by benchmarking the performance of nine deep neural architectures. We then segment the four subcortical structures: caudate, putamen, globus pallidus, and thalamus; white matter hyperintensities, and the normal appearing white matter. We show excellent generalizing capabilities across whole brain hemispheres in different specimens, and also on unseen images acquired at different magnetic field strengths and different imaging sequence. We then compute volumetric and localized cortical thickness measurements across key regions, and link them with semi-quantitative neuropathological ratings. Our code, containerized executables, and the processed datasets are publicly available at: https://github.com/Pulkit-Khandelwal/upenn-picsl-brain-ex-vivo.

翻译：离体脑部MRI相较于活体MRI在可视化和表征详细神经解剖结构方面具有显著优势，有助于将微尺度组织学研究与形态测量分析联系起来。然而，用于离体MRI脑图谱的自动分割方法尚未充分发展，这主要归因于标注数据集的有限可用性，以及扫描仪硬件和采集协议的异质性。本研究提出了一个包含37例离体尸检人脑组织标本的高分辨率数据集，这些标本在7T全身MRI扫描仪上采集。我们通过基准测试九种深度神经架构的性能，开发了用于分割大脑皮层的深度学习流程。随后对四种皮层下结构（尾状核、壳核、苍白球和丘脑）、白质高信号区以及正常表现白质进行分割。研究显示，该方法在不同标本的全脑半球中具有优秀的泛化能力，并且适用于在不同磁场强度和不同成像序列下获取的未见过图像。我们计算了关键区域的体积和局部皮层厚度测量值，并将其与半定量神经病理学评分相关联。我们的代码、容器化可执行文件以及处理后的数据集已在以下网址公开：https://github.com/Pulkit-Khandelwal/upenn-picsl-brain-ex-vivo。