Computer Aided Screening of Breast Cancer Based on the Processing of Mammograms

乳房X光检查被认为是筛查乳腺癌的最可靠的方式，并且使用乳房X光检查的筛查程序对于早期乳腺癌的检测已经变得流行。乳腺X线摄影计算机辅助诊断（CAD）系统有可能被用作第二阅读器，以提高大规模筛查的可靠性。乳腺X线摄影CAD系统的开发已成为医学图像处理领域的重要研究课题之一。乳腺X线摄影CAD系统一般由两个子系统组成，一个是肿块检测系统，另一个是微钙化簇检测系统。在本研究中，我们开发了新的方法来检测恶性肿瘤和簇状微钙化点，在肿瘤检测系统中，采用了一种称为虹膜滤波器的非线性滤波器来增强乳房X线照片上的圆形阴影。采用了一种新的肿瘤候选者边界估计方法和特征参数来提取肿瘤的边界特征。在聚类微钙化点检测系统中，提出了基于形态学的微钙化点检测方法。系统的性能已在国立癌症中心医院东院进行了评估，并以过去一年半门诊病人的乳腺X线照片作为未知样本。乳房X光检查次数为3812次。系统性能如下：恶性肿瘤检测系统的真阳性检出率为88%，平均每幅图像的假阳性数为1.34。簇状微钙化检测系统的准确率分别为94%和0.39。这些结果表明了所提出的系统的有效性。

Mammogram is considered to be the most reliable modality for the screening of breast cancer, and screening program using mammography has become popular for the detection of early breast cancer. Computer aided diagnosis (CAD) system for mammography has the possibility to be used as a second reader to increase reliability of mass screening. And development of CAD system for mammography has become one of the most important subjects of research in the area of medical image processing. CAD system for mammography consists of two subsystems in general.One of them is a tumor mass detection system and the other is a clustered microcalcification detection system. In this research, we have developed new methods to detect malignant tumors and clustered microcalcifications.In the tumor detection system, a nonlinear filter called Iris filter has been adopted to enhance rounded opacities on mammogram. A new boundary estimation method of tumor candidates and feature parameters to extract boundary characteristics of tumors have been adopted. In the clustered microcalcification detection system, morphological operation to detect microcalcifications has been developed. And adaptive processing is also adopted to keep the system optimum under various imaging conditions and patient variations.The system performance has been evaluated at National Cancer Center Hospital East.Mammograms taken from outpatients during the past one and a half years were used as unknown samples. The number of mammograms is 3812. The system performance was as follows.True positive detection rate and average number of false positives of the malignant tumor detection system were 88% and 1.34 per image, respectively. Those of the clustered microcalcification detection system were 94% and 0.39, respectively. These results show the effectiveness of proposed system..