基于一致性直方图的超声乳腺图片分割方法

为了在乳腺超声图像中准确的分割出病灶,提出了新的一致性定义,把纹理信息应用到计算一致性中.基于最大熵准则求一致性阈值,把超声图片分为两个子集,一致性区域和非一致性区域;对一致性区域采取求直方图谷值的方法分割,而用邻域信息处理非一致性区域,从而完成对图像的分割.实验结果表明:该分割过程既考虑了图像的局部信息,又考虑了全局信息,弥补了传统的直方图分割算法无法包含局部信息的缺陷.算法处理结果得到了超声医学专家的认可;通过采用差异实验方法评估,取得了理想的效果.

A method for breast ultrasound image segmentation based on homogeneity histogram

In order to detect potential lesions, a new method of breast ultrasound image segmentation based on the homogeneity histogram is proposed. Texture as well as edge features were used in the computation of homogeneity, so both global and local messages were considered, which can’t be achieved by previous algorithms. The images were divided into the homogeneity subset and the non-homogeneity subset according to the threshold computed from the maximum entropy principle. The two subsets were segmented separately thereafter. Radio experts affirmed the processed result, and the difference assessment experiment demonstrated that the proposed approach could generate valid nodules discriminating from breast sonographics with valuable TP, FP, and FN data. It will be a better assistant for radiologists in the diagnosis of breast cancer.