Department of Biomedical Engineering, University of Alberta, 1098 Research Transition Facility, 8308-114 Street, Edmonton, Alberta T6G 2V2, Canada.
Neuroimage. 2021 Aug 15;237:118105. doi: 10.1016/j.neuroimage.2021.118105. Epub 2021 Apr 29.
To extract Diffusion Tensor Imaging (DTI) parameters from the human cortex, the inner and outer boundaries of the cortex are usually defined on 3D-T1-weighted images and then applied to the co-registered DTI. However, this analysis requires the acquisition of an additional high-resolution structural image that may not be practical in various imaging studies. Here an automatic cortical boundary segmentation method was developed to work directly only on the native DTI images by using fractional anisotropy (FA) maps and mean diffusion weighted images (DWI), the latter with acceptable gray-white matter image contrast. This new method was compared to the conventional cortical segmentations generated from high-resolution T1 structural images in 5 participants. In addition, the proposed method was applied to 15 healthy young adults (10 cross-sectional, 5 test-retest) to measure FA, MD, and radiality of the primary eigenvector across the cortex on whole-brain 1.5 mm isotropic images acquired in 3.5 min at 3T. The proposed method generated reasonable segmentations of the cortical boundaries for all individuals and large proportions of the proposed method segmentations (more than 85%) were within ±1 mm from those generated with the conventional approach on higher resolution T1 structural images. Both FA (0.15) and MD (0.77 × 10 mm/s) extracted halfway between the cortical boundaries were relatively stable across the cortex, although focal regions such as the posterior bank of the central sulcus, anterior insula, and medial temporal lobe showed higher FA. The primary eigenvectors were primarily oriented radially to the middle cortical surface, but there were tangential orientations in the sulcal fundi as well as in the posterior bank of the central sulcus. The proposed method demonstrates the feasibility and accuracy of cortical analysis in native DTI space while avoiding the acquisition of other imaging contrasts like 3D T1-weighted scans.
为了从人脑皮质中提取弥散张量成像（DTI）参数，通常在 3D-T1 加权图像上定义皮质的内、外边界，然后将其应用于配准的 DTI。然而，这种分析需要获取额外的高分辨率结构图像，而在各种成像研究中可能并不实用。在这里，我们开发了一种自动皮质边界分割方法，该方法仅通过使用各向异性分数（FA）图和平均弥散加权图像（DWI）来工作，而后者具有可接受的灰白质图像对比度，后者无需额外的高分辨率结构图像。该新方法与在 5 名参与者中生成的传统皮质分割方法进行了比较。此外，该方法还应用于 15 名健康的年轻成年人（10 名横断面，5 名测试-再测试），以在 3T 上使用 3.5 分钟采集的 1.5mm 各向同性全脑图像测量整个皮质的主特征向量的 FA、MD 和放射率。该方法为所有人生成了合理的皮质边界分割，并且大部分（超过 85%）的方法分割与在更高分辨率 T1 结构图像上生成的传统方法分割的边界之间的距离在±1mm 以内。在整个皮质中，提取自皮质边界之间一半处的 FA（约 0.15）和 MD（约 0.77×10mm/s）都相对稳定，尽管像中央沟后缘、前岛叶和内侧颞叶等焦点区域的 FA 更高。主特征向量主要朝向皮质的中间表面呈放射状，但在脑回的底部以及中央沟的后缘也存在切向方向。该方法证明了在避免获取其他成像对比（如 3D T1 加权扫描）的情况下，在原始 DTI 空间中进行皮质分析的可行性和准确性。
