基于层间先验知识从脑MRI图像中自动提取脑组织

Inter-slice prior knowledge based automatic brain tissue extraction from magnetic resonance images of brain

目的从脑MRI图像中提取脑组织,解决边缘模糊时脑和非脑组织难以分离的问题。方法首先利用各向异性扩散滤波的方法对脑MRI图像进行去噪处理;然后利用形态学的方法对初始脑MRI图像进行脑组织提取,在此分割结果的基础上,利用相邻层脑形态差异较小的特点,实现结构元素的自适应选取,完成从脑MRI图像中逐层准确、自动提取脑组织。结果采用不同来源的数据对算法性能进行了测试,结果优于经典背散射电子成像(BSE)方法的分割结果。结论利用层间先验知识有利于实现边缘模糊的脑组织自动准确提取,且适用性较强。

Objective To extract brain tissue from magnetic resonance imaging（MRI） images,and solve the problem of fuzzy boundaries between the brain and non-brain tissues.Methods First,the MRI images were denoised by anisotropic diffusion filter and then the mathematical morphological methods were used to extract brain tissues from initial slice.Based on the initial results,the prior knowledge of small morphological variation in brain adjacent slice was used to adjust the adaptive selection of structure element and realize the accurate and automatic extraction of the brain tissues slice by slice.Results The performance of the proposed method was evaluated using different dataset,and the results were superior to those obtained from classical back scattered electron imaging（BSE） method.Conclusion It is demonstrated that application of inter-slice prior knowledge could be beneficial for the automatic brain extraction and the algorithm is accurate and generic.