Department of Clinical Radiology, University of Münster, Münster, Germany.
Clin Neuroradiol. 2012 Mar;22(1):79-91. doi: 10.1007/s00062-011-0123-0. Epub 2012 Jan 21.
The aim of this study was to present and evaluate a standardized technique for brain segmentation of cranial computed tomography (CT) using probabilistic partial volume tissue maps based on a database of high resolution T1 magnetic resonance images (MRI).
Probabilistic tissue maps of white matter (WM), gray matter (GM) and cerebrospinal fluid (CSF) were derived from 600 normal brain MRIs (3.0 Tesla, T1-3D-turbo-field-echo) of 2 large community-based population studies (BiDirect and SEARCH Health studies). After partial tissue segmentation (FAST 4.0), MR images were linearly registered to MNI-152 standard space (FLIRT 5.5) with non-linear refinement (FNIRT 1.0) to obtain non-binary probabilistic volume images for each tissue class which were subsequently used for CT segmentation. From 150 normal cerebral CT scans a customized reference image in standard space was constructed with iterative non-linear registration to MNI-152 space. The inverse warp of tissue-specific probability maps to CT space (MNI-152 to individual CT) was used to decompose a CT image into tissue specific components (GM, WM, CSF).
Potential benefits and utility of this novel approach with regard to unsupervised quantification of CT images and possible visual enhancement are addressed. Illustrative examples of tissue segmentation in different pathological cases including perfusion CT are presented.
Automated tissue segmentation of cranial CT images using highly refined tissue probability maps derived from high resolution MR images is feasible. Potential applications include automated quantification of WM in leukoaraiosis, CSF in hydrocephalic patients, GM in neurodegeneration and ischemia and perfusion maps with separate assessment of GM and WM.
本研究旨在介绍并评估一种基于高分辨率 T1 磁共振成像（MRI）数据库的基于概率部分容积组织图谱的颅 CT 脑分割标准化技术。
从 2 项大型社区人群研究（BiDirect 和 SEARCH Health 研究）中的 600 例正常脑 MRI（3.0T，T1-3D-turbo-field-echo）中得出白质（WM）、灰质（GM）和脑脊液（CSF）的概率组织图谱。在进行部分组织分割（FAST 4.0）后，将 MRI 线性注册到 MNI-152 标准空间（FLIRT 5.5），并进行非线性细化（FNIRT 1.0），以获得每个组织类别的非二进制概率体图像，随后用于 CT 分割。从 150 例正常脑 CT 扫描中，使用迭代非线性注册到 MNI-152 空间，构建了标准空间中的定制参考图像。组织特异性概率图到 CT 空间（MNI-152 到个体 CT）的逆变换用于将 CT 图像分解为组织特异性成分（GM、WM、CSF）。
本研究探讨了这种新方法在 CT 图像的无监督定量分析和可能的视觉增强方面的潜在优势和实用性。本文还展示了不同病理情况下（包括灌注 CT）的组织分割示例。
使用从高分辨率 MRI 中提取的高度细化的组织概率图谱自动分割颅 CT 图像是可行的。潜在的应用包括在白质疏松症中对 WM 进行自动定量分析、在脑积水患者中对 CSF 进行自动定量分析、在神经退行性变和缺血中对 GM 进行自动定量分析以及对 GM 和 WM 进行单独评估的灌注图。
