中文摘要：

乳腺癌是妇女常见恶性肿瘤之一,早期诊断和治疗是降低乳腺癌患者死亡率的关键。微钙化是乳腺癌早期的一个重要标志,因此快速准确地找出乳腺X光片中含有微钙化簇的感兴趣区域（ROI）是成功诊断的第一步。乳腺X光片中含有大量无病变区域和少量微钙化区域,形成了一种典型的不对称分类问题。本研究结合大量无病变区域的信息训练多级组合分类器,并借助多尺度方法加快筛选速度,以定位ROI。在真实的数字化X线乳腺照片上的实验表明,该方法在无漏检的情况下,可以排除92.64%的正常区域,而且基于Matlab处理,对于每幅图片的平均处理时间仅为7 s。

英文摘要：

Breast cancer is one of the common malignant tumors for women. The key to reducing the mortality caused by this disease lies in the early detection and treatment. Microcalcification is an important sign of the breast cancer in the early stage. Therefore the first step of successful diagnosis is to locate the region containing microcalcification cluster in the mammograms. It becomes a typical issue of asymmetry because the large portion of the mammograms is the normal region while only a small portion is the microcalcification region. A cascade assembled classifier was trained to locate the region of interest （ROI） with the abundant information from the normal region. By the muhi-scale processing method, the screening speed increased. The experimental results on the mammograms indicated that this method could exclude 92.64 % of normal area and the average time used for detecting each image is only 7 s based on Matlab.