基于支持向量机的乳腺钙化点检测算法

Micro-calcifications detection algorithm in mammogram based on support vector machine

针对乳腺X线图像微钙化点检测假阳性高的问题,提出一种微钙化点检测算法。算法首先以小波与Top-hat算子相结合的方法进行钙化点粗检测,然后以支持向量机(SVM:Support Vector Machine)为工具对粗检测结果进行真钙化点与假钙化点分类。对开放乳腺图像数据库MIAS的仿真实验表明,算法的检出率超过98%,错检率不足4%,达到理想的检测效果。

To solve the problem of high false positive detection in micro-calcifications of mammogram, this paper presented a detective algorithm on micro-calcifications. The algorithm first used wavelet and Top-hat to coarse detection on micro-calcifications, and then distributed true or false calcifications on the results of coarse detect with SVM. According to the mammogram database MIAS simulation experiment, the rate of detection calcifications was more than 98%, while the false positive was less than 4% and the algorithm is effectiveness.