RF01-05 Topological Analysis of Clusters of Microcalcifications in Digital Mammography Details

Start Date: 2025-06-25

End Date: 2025-06-25

This research addresses how to process mammograms to detect clusters of microcalcifications while minimizing false negatives (missed clusters) and false positives (detected but nonexistent clusters). The system confirms whether a finding corresponds to a calcification using a classifier trained on extracted features. Two methods are proposed: one for identifying individual calcifications and another for detecting clusters. A cluster is defined as three or more microcalcifications within a 1 cm² area, with detection prioritizing regions of higher density. The system’s output is a list of centroids for calcifications and clusters, allowing calculation of the number of clusters per mammogram and the number of calcifications in each. The key contribution lies in distinguishing true findings from false ones, thereby improving cluster detection and diagnostic accuracy.