Mammographic parenchymal texture analysis for estrogen-receptor subtype specific breast cancer risk estimation

Abstract

We investigate the potential of mammographic parenchymal texture as a surrogate marker of the risk to develop Estrogen Receptor (ER) sub-type specific breast cancer. A case-control study was performed, including 118 cancer cases stratified by ER receptor status and 354 age-matched controls. Digital mammographic (DM) images were retrospectively collected and analyzed under HIPAA and IRB approval. The performance of the texture features was compared to that of the standard mammographic density measures. We observed that breast percent density PD% and parenchymal texture features can both distinguish between cancer cases and controls (A z > 0.70). However, for ER subtype-specific classification, PD% alone does not provide sufficient classification (A z = 0.60), while texture features have significant classification performance (A z = 0.70). Combining breast density with texture features achieves the best performance (A z = 0.71). These findings suggest that mammographic texture analysis may have value for sub-type specific breast cancer risk assessment. © 2012 Springer-Verlag Berlin Heidelberg.