In addition to the DDSM data we have collected our own mammography database containing 961 mass regions at a large radiological center, which we will make publicly available at the well-known UCI machine learning repository17 with the publication of this paper. The database is based on modern full-field digital mammograms and contains cases that have been acquired from 2003 to 2006. We will refer to this set of mass regions as MUCI. 515 53.6% of these mass regions are benign and 446 46.4% are malignant. Similar to the DDSM data we use the patient’s age and BI-RADS descriptions as input attributes for the CBR system. In addition to the BI-RADS descriptions mentioned earlier, the mass density was available as a BI-RADS compliant attribute for these cases and we therefore use it as an additional input attribute.