基于自动随机游走的乳腺肿块分割算法

针对乳腺X线影像肿块分割易受弱边缘和周围组织干扰的问题,提出一种基于自动随机游走的乳腺肿块分割算法.利用二维最大熵阈值法、区域生长及形态学方法自动确定一系列标记点,采用平均边缘梯度评价法选择有效标记点进行随机游走分割以获得初步分割结果,并在此分割基础上进行星芒状结构检测,获得完整的肿块分割边缘.随机选取227例肿块图像进行分割,对分割结果进行特征提取和分类.实验结果表明,该算法克服了半自动随机游走的应用局限性,提高了乳腺肿块的分割精度；与其他分割算法相比,该算法在后续的分类中具有更高的分类精度.

Mammographic mass segmentation algorithm based on automatic random walks

A mammographic mass segmentation algorithm based on automatic random walks algorithm was presented in order to overcome the interference of the weak edge and surrounding tissues on mass segmentation in mammograms. Two dimensional maximum entropy threshold, region growing algorithm and morphological method were used to automatically get a series of labels. Then the evaluation method of average edge gradient was used to select the effective labels for random walks segmentation, and the initial segmentation results were obtained; the spiculation pattern was also detected based on such segmentation results. The complete segmentation contour for mass was achieved. 227 images containing mass were randomly selected for segmentation. Then feature extraction and classification were implemented based on the segmentation results. Experimental results show that the algorithm overcomes the application limitation of semi-automatic random walks algorithm, and improves the accuracy of segmentation. The algorithm achieves higher classification accuracy compared with other segmentation algorithms for mass.