Adamson Chris, Beare Richard, Walterfang Mark, Seal Marc
Royal Childrens Hospital, 50 Flemington Road Parkville, VIC, Melbourne, 3052, Australia,
Neuroinformatics. 2014 Oct;12(4):595-614. doi: 10.1007/s12021-014-9236-3.
This paper presents a fully automated pipeline for thickness profile evaluation and analysis of the human corpus callosum (CC) in 3D structural T 1-weighted magnetic resonance images. The pipeline performs the following sequence of steps: midsagittal plane extraction, CC segmentation algorithm, quality control tool, thickness profile generation, statistical analysis and results figure generator. The CC segmentation algorithm is a novel technique that is based on a template-based initialisation with refinement using mathematical morphology operations. The algorithm is demonstrated to have high segmentation accuracy when compared to manual segmentations on two large, publicly available datasets. Additionally, the resultant thickness profiles generated from the automated segmentations are shown to be highly correlated to those generated from the ground truth segmentations. The manual editing tool provides a user-friendly environment for correction of errors and quality control. Statistical analysis and a novel figure generator are provided to facilitate group-wise morphological analysis of the CC.
本文提出了一种全自动流程，用于在三维结构T1加权磁共振图像中评估和分析人类胼胝体（CC）的厚度轮廓。该流程执行以下一系列步骤：正中矢状面提取、CC分割算法、质量控制工具、厚度轮廓生成、统计分析和结果图生成器。CC分割算法是一种新技术，它基于基于模板的初始化，并使用数学形态学操作进行细化。与在两个大型公开可用数据集上的手动分割相比，该算法具有很高的分割精度。此外，自动分割生成的厚度轮廓与真实分割生成的厚度轮廓高度相关。手动编辑工具为错误校正和质量控制提供了用户友好的环境。提供统计分析和新颖的数据图生成器，以促进对CC进行分组形态分析。
