Semi-automated and fully automated mammographic density measurement and breast cancer risk prediction

摘要: The task of breast density quantification is becoming increasingly relevant due to its association with cancer risk. In this work, a semi-automated and fully automated tools assess from full-field digitized mammograms are presented. first tool based on supervised interactive thresholding procedure for segmenting dense fatty tissue used twofold goal: assessing mammographic (MD) in more objective accurate way than via visual-based methods labeling the that later employed train tool. Although most rely approaches global mammogram, proposed method relies pixel-level labeling, allowing better classification measurement continuous scale. presented combines scheme local features operations improve performance classifier. A dataset 655 was test concordance both measuring MD. Three expert radiologists measured MD each using (DM-Scan). It then by system correlation between computed. relation analyzed case-control consisting 230 mammograms. Intraclass Correlation Coefficient (ICC) compute reliability among raters techniques. results obtained showed an average ICC=0.922 when tool, whilst measures ICC=0.838. study, Odds Ratios (OR) 1.38 1.50 per 10% increase respectively. can therefore be concluded assessments present good correlation. Both also found risk, which warrants risk prediction clinical decision making. full version DM-Scan freely available.