Abstract: Microcalcification clusters(MCs)are important signs for early breast cancer detection. The existing initial detection methods result in lots of false positive data because of the requirements of high sensitivity. A classification strategy combining with global and local views was proposed based on MCs characteristics. The costsensitive SVM classification algorithm was employed according to different misclassification costs of true and false positive instances. In the construction of classification model， the parameters and feature subset were optimized with particle swarm method to enhance the generalization performance. The ensemble method reduces excessive false positive data but with high sensitivity， and removes redundant and irrelevant features. Experimental results show that the proposed method improves the performance of traditional methods on microcalcification clusters recognition.

Key words: microcalcification cluster detection, computeraided diagnosis, costsensitive learning, ensemble classification, particle swarm optimization, feature selection