基于数据融合的脑组织图像分割

Multi-spectral MR Image Segmentation of Brain

随着医学影像技术的发展,我们可以用不同的成像方法对同一个脑断层得到多模态的核磁共振图像,针对脑组织分割的需要,文中介绍了一种基于数据融合的多模分割方法。算法首先用基于模糊C均值聚类(FuzzyC-Means,FCM)的方法分别对单一模态的图像聚类进行分割,然后采用数据融合的方法得出最终的分割结果。实验结果表明,此方法能有效地分割出白质、灰质和脑脊液,并且分割精度要明显高于对单一模态图像的分割结果。

A novel multi-spectral MR image segmentation method based on multi-modality image fusion is presented in this paper. The basic idea of the method is that every single image is clustered using the FCM clustering method first, and then the final clustering results can be obtained by data fusion. The experimental results show that the presented method separates the white matter, gray matter and cerebro-spinal fluid （CSF） more accurately than does any single image segmentation method.