A new scheme for the computer-aided diagnosis (CAD) of microcalcification clusters (MCCs) detection in a Multi-Instance Learning (MIL) framework is proposed in this paper. To achieve a satisfactory performance, our algorithm first searches for possible candidates of microcalcification clusters using the mean-shift algorithm. Then, features are extracted from the potential candidates based on a constructed graph. Finally, a multi-instance learning method which locates the key instance in each bag of features is used to classify the possible candidates. Experimental results show that our scheme can achieve a superior performance on public datasets, and the computation is efficient.
展开▼