摘要
在乳腺图像中,肿块大多被埋没在复杂的、高密度的腺体背景中难以检测。针对这一问题,提出了一种基于金字塔结构的乳腺肿块自动检测方法。文中对几种典型的金字塔结构的构造方法做了比较;提出了一种使用BP人工神经网络用于实现低分辨率图像中肿块种子区域检测的新方法;提出了一种新的权值差别规则,同时添加了标志锥,使得生长算法不再严格受限于肿块种子的面积和形状。实验结果证明这种方法对于辅助临床医生诊断乳腺病变是有效的。
Lesions are usually difficult to detect as they often superimpose on dense structured background. In order to solve this problem, an approach of auto-detection of lesions based on image pyramid is presented in this paper. In the lower resolution image,a mass seed-region was detected by an ANN. With an improved grow tree algorithm, the edge of lesions was refined in the higher resolution image. A label pyramid is proposed in this method to make the tree grow method never restricted by area and shape of the seed-region. Experimental results show that the proposed approach is applicable to assist the radiologist's diagnosis.
出处
《上海生物医学工程》
2002年第2期8-11,共4页
Shanghai Journal of Biomedical Engineering
