Detecting Clusters of Microcalcifications with a Cascade-Based Approach

Abstract

In this paper we present a cascade-based framework to detect clusters of microcalcifications on mammograms. The algorithm is based on a sliding window technique where a detector is structured as a “cascade” of simple boosting classifiers with increasing complexity. Such a method couples the effectiveness of the cascade approach with the Rank- Boost algorithm that is aimed at maximizing the area under the ROC curve and represents a good choice when dealing with unbalanced data sets.