PDF(Pubmed)
Abstract：
The claustrum has a unique thin sheet-like structure that makes it hard to identify in typical anatomical MRI scans. Attempts have been made to identify the claustrum in anatomical images with either automatic segmentation techniques or using atlas-based approaches. However, the resulting labels fail to include the ventral claustrum portion, which consists of fragmented grey matter referred to as \"puddles\". The current dataset is a high-resolution label of the whole claustrum manually defined using an ultra-high resolution postmortem MRI image of one individual. Manual labelling was performed by four independent research trainees. Two trainees labelled the left claustrum and another two trainees labelled the right claustrum. For every hemisphere we created a union of the two labels and assessed the label correspondence using dice coefficients. We provide size measurements of the labels in MNI space by calculating the oriented bounding box size. These data are the first manual claustrum segmentation labels that include both the dorsal and ventral claustrum regions at such a high resolution in standard space. The label can be used to approximate the claustrum location in typical in vivo MRI scans of healthy individuals.
摘要：
锁骨具有独特的薄片状结构，这使得在典型的解剖MRI扫描中很难识别。已经尝试通过自动分割技术或使用基于图谱的方法来识别解剖图像中的锁骨。然而,由此产生的标签不包括腹侧锁骨部分，它由被称为“水坑”的碎片灰质组成。当前数据集是使用一个个体的超高分辨率死后MRI图像手动定义的整个锁骨的高分辨率标签。手动标签由四名独立研究学员进行。两名受训者标记为左肩带，另外两名受训者标记为右肩带。对于每个半球，我们创建了两个标签的并集，并使用骰子系数评估了标签对应关系。我们通过计算定向边界框大小来提供MNI空间中标签的大小测量。这些数据是在标准空间中以如此高的分辨率包括背侧和腹侧锁骨区域的第一手动锁骨分割标签。该标记可用于近似健康个体的典型体内MRI扫描中的锁骨位置。
