This abstract presents a novel classification approach for microcalcifications extracted from core biopsy tissue samples digitized using microCT. Because of lack of ground truth, it is investigated whether the application of a clustering method as a preprocessing step before classification would improve the obtained classification results.