基于细胞自动机的MRI脑肿瘤分割方法

提出一种基于细胞自动机的脑肿瘤分割方法.首先通过人工交互输入一条线,使用细胞自动机模型对脑肿瘤图像进行分割,得到肿瘤图像的标号图,然后使用活动轮廓模型对标号图进行优化处理,除去非肿瘤像素点的干扰,得到更平滑的脑肿瘤轮廓.使用该方法在对比增强T1加权MRI脑肿瘤图像进行分割实验.实验结果表明,此方法能够很好地解决脑肿瘤分割过程中容易出现的不完全分割问题,分割准确率(Dice相似系数)可达到(94.07±1.58)％.

Segmentation of Brain Tumors on MRI Images Based on Cellular Automata

An approach for brain tumor segmentation based on cellular automata is proposed. Cellular automata model is adopted to obtain the label map of brain tumor using the interactive user input (a line). Then, active contour model is used to refine the labeled map, which can eliminate the interference from non-tumor pixels and get more smooth contour of brain tumors. The proposed approach is applied to segment the brain tumor in the contrast enhanced T1 weighted magnetic resonance (MR) images. The experiment results show that the proposed approach can effectively solve the problem of incomplete segmentation which may occur in the segmentation results by the other segmentation method, and the segmentation accuracy (Dice similarity coefficient) can reach 94.07±1.58%.