Assessing Breast Cancer Masking Risk with Automated Texture Analysis in Full Field Digital Mammography

M. Kallenberg, M. Lillholm, P. Diao, K. Petersen, K. Holland, N. Karssemeijer, C. Igel and M. Nielsen

Annual Meeting of the Radiological Society of North America 2015.

Cited by ~3

PURPOSE The goal of this work is to develop a method to assess the risk of breast cancer masking, based on image characteristics beyond breast density. METHOD AND MATERIALS From the Dutch breast cancer screening program we collected 285 screen detected cancers, and 109 cancers that were screen negative and subsequently appeared as interval cancers. To obtain mammograms without cancerous tissue, we took the contralateral mammograms. We developed a novel machine learning based method called convolutional sparse autoencoder to characterize mammographic texture. The reason for focusing on mammographic texture rather than the amount of breast density is that a developing cancer may not only be masked because it is obscured; it may also be masked because its mammographic signs resemble the texture of normal tissue. The method was trained and tested on raw mammograms to determine cancer detection status in a five-fold cross validation. To assess the interaction of the texture scores with breast density, Volpara Density Grade (VDG) was determined for each image using Volpara, Matakina Technology, New Zealand. RESULTS We grouped women into low (VDG 1/2) versus high (VDG 3/4) dense, and low (Quartile 1/2) versus high (Quartile 3/4) texture risk score. We computed odds ratios (OR) for breast cancer masking risk (i.e. interval versus screen detected cancer) for each of the subgroups. The OR was 1.63 (1.04-2.53 95%CI) for the high dense group (as compared to the low dense group), whereas for the high texture score group (as compared to the low texture score group) this OR was 2.19 (1.37-3.49). Women who were classified as low dense but had a high texture score had a higher masking risk (OR 1.66 (0.53-5.20)) than women with dense breasts but a low texture score. CONCLUSION Mammographic texture is associated with breast cancer masking risk. We were able to identify a subgroup of women who are at an increased risk of having a cancer that is not detected due to textural masking, even though their breasts are non-dense. CLINICAL RELEVANCE/APPLICATION Automatic texture analysis enables assessing the risk that a breast cancer is masked in regular mammography, independently of breast density. As such it offers opportunities to further enhance personalized breast cancer screening, beyond breast density.