Novel multiresolution mammographic density segmentation using pseudo 3D features and adaptive cluster merging
Wenda He, Arne Juette, Erica R. E. Denton, Reyer Zwiggelaar · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2015
ABSTRACT Breast cancer is the most frequently diagnosed cancer in women. Early detection, precise identi cation of womenat risk, and application of appropriate disease prevention measures are by far the most e ective ways to overcomethe disease. Successful mammographic density segmentation is a key aspect in deriving correct tissue composition,ensuring an accurate mammographic risk assessment. However, mammographic densities have not yet been fullyincorporated with non-image based risk prediction models, ( e.g. the Gail and the Tyrer-Cuzick model), becauseof unreliable segmentation consistency and accuracy. This paper presents a novel multiresolution mammographicdensity segmentation, a concept of stack representation is proposed, and 3D texture features were extracted byadapting techniques based on classic 2D rst-order statistics. An unsupervised clustering technique was employedto achieve mammographic segmentation, in which two improvements were made; 1) consistent segmentation byincorporating an optimal centroids initialisation step, and 2) signi cantly reduced the number of missegmentationby using an adaptive cluster merging technique. A set of full eld digital mammograms was used in the evaluation.Visual assessment indicated substantial improvement on segmented anatomical structures and tissue speci careas, especially in low mammographic density categories. The developed method demonstrated an ability toimprove the quality of mammographic segmentation via clustering, and results indicated an improvement of 26%in segmented image with good quality when compared with the standard clustering approach. This in turn canbe found useful in early breast cancer detection, risk-strati ed screening, and aiding radiologists in the processof decision making prior to surgery and/or treatment.Keywords: BI-RADS, computer aided mammography, mammographic density segmentation