Influence of segmentation on classification of microcalcifications in digital mammography

Contrast of microcalcifications can be used to classify benign and malignant types. Different measures for contrast are investigated: mean and maximum contrast, with and without correction for microcalcification size. It is analyzed how the discriminating power of contrast depends on the segmentation process. For classification the k-Nearest-Neighbor method is used and for testing the “leave-one-out-method”. Results of an experimental study using a dataset of mammographic images digitized at 2048×2048 are presented. It is shown that segmentation strongly influences classification