脑血肿CT图像分割的改进快速行进方法

Improved fast marching method for CT image segmentation of brain hematoma

为解决脑血肿CT图像的手动分割和半自动分割方法在手术计划制定过程中存在分割速度无法满足临床时间要求的问题,对快速行进(fast marching,FM)方法进行改进,提出一种三维全自动脑血肿CT图像分割方法。利用阈值处理自动获取种子点集,解决手动设置种子点的问题,实现全自动;提取感兴趣区域减少计算量;通过感兴趣区域金字塔的迭代分割进一步减少FM方法的演化计算量,使分割结果快速向边界收敛。实验结果表明,该方法能够自动、准确并高效地分割脑血肿CT图像。

To solve the problem that the manual and semi-automatic segmentation methods of CT images of cerebral hematoma can not meet the clinical time requirements in the surgical planning process,the fast marching(FM)method was improved,and a three-dimensional automatic segmentation method of CT images of cerebral hematoma was proposed.Automatic acquisition of seed point set by threshold processing solved the problem of setting seed point manually and realized full automation.The amount of data to be calculated was reduced by extracting the regions of interest.The iterative segmentation of pyramid of regions of interest further reduced the evolutionary computation of FM method,and made the segmentation result converge to the boundary quickly.Experimental results show that the proposed method can automatically,accurately and efficiently segment CT images of cerebral hematoma.