Mammographic parenchymal pattern segmentation: A clinical evaluation

Mammographic risk assessment is becoming more important in decision making within screening mammography and computer aided diagnosis. Strong evidence shows that characteristic mixtures of breast tissues as seen on mammography, referred to as mammographic parenchymal patterns, provide crucial information about breast cancer risk. One approach to automatic mammographic risk assessment concentrates on mammographic parenchymal pattern segmentation, to quantify the relative proportion of different breast tissues. This paper presents a clinical evaluation of mammographic parenchymal pattern segmentation based on Tabár's tissue modelling, using the Mammographic Image Analysis Society (MIAS) database. The segmentation assessment results show a strong correlation with increasing Tabár and Birads risk categories. In addition, the segmentation assessment is linked to the correct/incorrect automatic mammographic risk classification, which indicated that good segmentation results tend to lead to correct mammographic risk estimation.