基于二维最大熵阈值分割的钙化点检测算法

A Calcification Detection Method Based on Two-Dimensional Entropic Thresholding

在数字乳腺X线图像中,钙化是早期乳腺癌的重要征象之一。为了提高钙化点检测的准确度及降低检测的假阳性率,提出了一种结合数学形态学滤波和二维最大熵阈值分割的钙化点检测算法。算法首先采用top-hat算子对图像的背景进行抑制,然后利用二维最大熵阈值分割算法得到可疑钙化点区域,最后采用SVM分类的方法去除假阳性区域,得到最终的钙化点检测结果,并采用MIAS乳腺影像库进行仿真实验,钙化点检测的敏感性为94.6%,假阳性率为10.5%。实验结果表明,方法对钙化点的定位精确,具有较高的检出率及较低的假阳性率。

Microcalcifications are the primary radiological symptom of early breast cancer in digital mammography. In order to improve the accuracy and reduce the false positive rate of calcification detection,this paper proposed an algorithm for automatic detection of calcification in digital mammograms. The method first restrains the background based on mathematical morphology filter named Top-hat operator,then adopts the two-dimensional entropic thresholding method to get the calcification doubt regions. Finally,Support Vector Machine（SVM） is used to remove the false positive regions. Experiment on the MIAS（Mammographic Image Analysis Society） database shows that our method can detect the calcifications accurately,with high sensitivity（94.6%）and low false positive rate（10.5%）.