Detection of microcalcifications in digital mammograms.

A method is described for processing digital mammograms to detect microcalcifications that, although they may be so small that they are almost undetectable visually, could be indicators of a possible malignancy. The process consists of three stages. In the first, zones of the mammogram that potentially correspond to microcalcifications are extracted. This is done by analyzing the distribution of brightness over the entire mammogram, seeking structures that are typical of microcalcifications. The resulting images are passed to the second stage, in which localized clusters are identified as regions-of-interest (ROI). The third stage retrieves any information that might have been lost in the previous stages. The end result is a mammogram in which the few ROIs that reached the final phase -- and which, in reality, are those that could be of true medical interest -- clearly stand out. The system was tested against the DDSM mammogram database. All of the malignant and benign lesions were located perfectly, as well as some difficult-to-classify calcified formations. The system is now fully operational, accepting, as well as the database images, both digital and analogical mammograms.