BASCNet: Bilateral adaptive spatial and channel attention network for breast density classification in the mammogram. (September 2021)

• Record Type: Journal Article
• Authors: Zhao, Wenwei; Wang, Runze; Qi, Yunliang; Lou, Meng; Wang, Yiming; Yang, Yang; Deng, Xiangyu; Ma, Yide
• Is Part Of: Biomedical signal processing and control. Volume 70(2021)
• Publication Date: 2021-09
• Subjects: Mammographic density -- Classification -- Adaptive spatial attention -- Adaptive channel attention; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.103073
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: An automatic end-to-end convolutional neural network model is designed for breast density classification in mammograms. By simulating the doctor's reading mechanism, we combine the information of bilateral breasts to classify breast density. Adaptive spatial attention module (ASAM) and adaptive channel attention module (ACAM), are employed to explore discriminant information for breast density classification. Proposed BASCNet has been verified on the DDSM and INbreast datasets, and both have achieved state-of-the-art results. Abstract: Breast density is a significant element for breast cancer precaution. The existing mammographic density classification methods cannot achieve satisfactory classification accuracy while achieving end-to-end. In this paper, we present a novel bilateral adaptive spatial and channel attention network (BASCNet) which integrates the information of the left and right breasts and adaptively pays attention to the discriminative features in spatial and channel dimensions. The proposed BASCNet has been fully proved on the public Digital Database for Screening Mammography (DDSM) and INbreast dataset, and the classification accuracies of 85.10% and 90.51% were achieved with fivefold cross-validation, respectively. Our method is fully automatic and has achieved the classification performance superior to the existing breast density classification methods. Massive ablation experiments were conducted to demonstrate the effectiveness of the network structure. Moreover, we compared the effects of different views (CC and MLO) on breast density classification and verified the effectiveness of the contralateral breast information integration. Overall, the proposed BASCNet has the potential to be applied to clinical diagnosis.