改进的Brain Extraction Tool算法及其在脑实质分割中的应用

Improved brain extraction tool algorithm for brain parenchyma segmentation

BET(Brain Extraction Tool)算法是一种常用的从磁共振(MRI)脑图像中分割脑实质的工具,在实际应用中发现,BET算法对正常脑实质的分割精度较高,但对有病灶的脑实质分割精度较差。根据BET算法存在的问题,改进原BET算法中不合理的u_3,简化了计算繁琐的u_2,并将其应用于分割MRI图像中的脑实质。首先:选择序列图像中间层,对其应用两次改进后的BET算法获得精确分割结果;然后:将获得的边界向其中心缩小一定比例后作为与其相邻层的初始边界再次应用修改后的算法获得该层精确边界;最后,不断重复上述步骤直至所有层分割结束。改进后的算法对脑部图像分割结果与人工分割结果的重叠率达到92.92%,而使用FSL中提供的BET工具的分割结果与人工分割结果的重叠率为88.94%。改进后的算法相比原BET算法能够更加准确地分割MRI图像中的脑实质。

Brain extraction tool （BET） algorithm is a commonly used tool for segmenting brain parenchyma from magnetic resonance image （MRI）. The practical application shows BET algorithm has a higher segmentation precision for the normal brain parenehyma than the brain parenchyma with lesions. According to the existing problems of BET algorithm, the unreasonable force u3 in original BET algorithm was revised, and the force u2 with tedious calculation was simplified. The modified BET algorithm was applied to segment the brain parenchyma of MR/. The improved BET algorithm was firstly applied to the selected intermediate slice of sequence images to obtain the accurate segmentation results. The obtained boundary was narrowed to its center in accordance with a certain proportion. The narrowed boundary was taken as the original boundary of the adjacent slices, and the improved BET algorithm was used again to obtain the accurate boundary of the slice. The above procedure was repeated until all slices were segmented. The overlap rate between the segmentation results of improved algorithm and the results of artificial segmentation reached 92.92%, while the overlap rate between the segmentation results of BET provided by FSL and the results of artificial segmentation reached 88.94%. Compared with the original BET algorithm, the improved algorithm can segment the brain parenchyma from MR/more accurately.