Abstract: An important feature of early breast cancer is microcalcification.The first step of successful diagnosis is how to identify microcalcifications rapidly and accurately in a mammogram.A new rapid detection method based on priori model and region growing is presented in this paper.A 0.5 mm diameter template is chosen to find out local peak points in a mammogram.The region growing method is used by taking these points as initial seeds.Features of each area as size，average gray and contrast are computed，and the areas whose features meeting characters of microcalcification are reserved.Based on priori knowledge，the microcalcification cluster is detected.Experiments show that microcalcifications can be detected by this algorithm rapidly and automatically，improving doctors’ diagnosis.

Key words: mammogram, microcalcification, region growing, priori model

摘要： 早期乳腺癌的一个重要特征就是钙化点，快速准确地找出乳腺图像中的钙化点是成功诊断的第一步。提出了一种先验模板和区域生长的钙化点快速检测方法。根据钙化点检测的临床经验，选用一直径为0.5 mm的模板找出乳腺图像中的局部峰值点。以这些峰值点为初始种子点，进行区域生长;计算每个区域的面积、平均灰度、对比度，保留满足钙化点特征的区域。根据先验知识，对生长获得的钙化点是否成簇进行判别，保留成簇的微钙化点。实验表明，该算法实现了乳腺图像中钙化点的快速自动检测，提高了医生诊断的正确性。

关键词: 乳腺图像, 微钙化点, 区域生长, 先验模型