基于帧间相关性的乳腺MRI三维分割

发布时间：2018-06-20 02:19

  本文选题：乳腺磁共振图像 + 病灶分割　； 参考：《天津大学学报(自然科学与工程技术版)》2017年08期

【摘要】：针对乳腺磁共振图像序列的肿瘤分割问题,提出一种基于超像素和改进C-V模型的三维全自动分割方法.该方法利用磁共振图像序列的帧间相关性,约束相邻帧图像的分割轮廓.采用超像素算法提取肿瘤的大致轮廓,再用改进的C-V水平集算法对可疑区域边缘进行优化,使其更接近肿瘤的实际边缘.将该方法及3种对比方法应用于89例乳腺MRI序列图像.以手动分割的轮廓为基准,该方法得到的平均重叠率为87.84%,,相比于C-V模型的58.90%,、超像素和水平集结合的76.36%,、K均值+C-V的83.62%,,有明显提升.实验结果表明,该方法的全自动分割结果对于肿瘤起始和终止帧图像具有较高的分割精度.
[Abstract]:A three dimensional fully automatic segmentation method based on super pixel and improved C-V model is proposed for tumor segmentation in the breast MRI sequence. This method uses the interframe correlation of the MRI image sequence to restrict the segmentation contour of the adjacent frame images. The rough contour of the tumor is extracted with the super pixel algorithm, and the improved C-V level set is used. The algorithm optimizes the edge of the suspicious area to make it closer to the actual edge of the tumor. This method and 3 contrast methods are applied to 89 cases of breast MRI sequences. The average overlap rate of this method is 87.84%, compared with 58.90% of the C-V model, 76.36% of the combination of super pixels and the level set, and the K mean +C- The experimental results show that the automatic segmentation results of V have a higher segmentation accuracy for tumor initiation and termination frame images. 83.62%.
【作者单位】： 天津大学电气自动化与信息工程学院;
【基金】：国家自然科学基金资助项目(61271069)~~
【分类号】：R737.9;TP391.41
，

本文编号：2042482