Search

We investigate various channel encoding techniques applied to breast density classification in mammograms; specifically, local binary, ternary, and quinary encoding approaches are considered. Subsequently, we propose a new encoding approach based on a seven-encoding technique, yielding a new local pattern operator called a local septenary pattern operator. Experimental results suggest that the proposed local pattern operator is robust and outperforms the other encoding techniques when evaluated on the Mammographic Image Analysis Society (MIAS) and InBreast datasets. The local septenary pattern operator achieved a maximum classification accuracy of 83.3% and 80.5% on the MIAS and InBreast datasets, respectively. The closest comparison achieved by the other local pattern operators is the local quinary operator, with maximum accuracies of 82.1% (MIAS) and 80.1% (InBreast), respectively., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2020. Published by Elsevier Ltd.)

This paper presents a method for automatic breast pectoral muscle segmentation in mediolateral oblique mammograms using a Convolutional Neural Network (CNN) inspired by the Holistically-nested Edge Detection (HED) network. Most of the existing methods in the literature are based on hand-crafted models such as straight-line, curve-based techniques or a combination of both. Unfortunately, such models are insufficient when dealing with complex shape variations of the pectoral muscle boundary and when the boundary is unclear due to overlapping breast tissue. To compensate for these issues, we propose a neural network framework that incorporates multi-scale and multi-level learning, capable of learning complex hierarchical features to resolve spatial ambiguity in estimating the pectoral muscle boundary. For this purpose, we modified the HED network architecture to specifically find 'contour-like' objects in mammograms. The proposed framework produced a probability map that can be used to estimate the initial pectoral muscle boundary. Subsequently, we process these maps by extracting morphological properties to find the actual pectoral muscle boundary. Finally, we developed two different post-processing steps to find the actual pectoral muscle boundary. Quantitative evaluation results show that the proposed method is comparable with alternative state-of-the-art methods producing on average values of 94.8 ± 8.5% and 97.5 ± 6.3% for the Jaccard and Dice similarity metrics, respectively, across four different databases., (Copyright © 2019 Elsevier B.V. All rights reserved.)