针对胼胝体的图像特点以及实际应用要求,采用半自动方法对MRI中的胼胝体进行分割。首先采用基于Live-Wire的算法对胼胝体影像的起始层和终止层进行初始分割,然后利用基于距离变换的形状插值算法获取中间层的初始轮廓信息,对插值获得的初始轮廓采用Snake模型进行局部收缩,获得真实的胼胝体边界。对序列MRI脑影像中的胼胝体进行分割、重建、标定。实验结果与临床医师的使用反馈证明,本文提出的算法具有较高的灵活性与可信度,对胼胝体的分割精度与解剖统计信息相符,分割结果可满足临床需求。
Based on the features of MR images of callosum and the requirements of practical application, a semi-automatic algorithm for callosum segmentation was proposed in this paper. Firstly, Live-Wire method was adopted to get the initial segment result for the first and last slices. Then the initial contour of middle slices was got using the method of shape interpolation based on distance transform. Using the initial contour as input of Snake model, the real callosum boundary could be found after some times of iteration. The experimental results and the feedbacks of clinical doctor showed that the algorithm has good flexibility and credibility, and the segment results match the Atlas of anatomy and meet the clinical needs.