Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2020 Jan;47(1):110-118.
doi: 10.1002/mp.13886. Epub 2019 Nov 19.

Deep learning modeling using normal mammograms for predicting breast cancer risk

Dooman Arefan  1 , Aly A Mohamed  1 , Wendie A Berg  1   2 , Margarita L Zuley  1   2 , Jules H Sumkin  1   2 , Shandong Wu  3
Affiliations

Deep learning modeling using normal mammograms for predicting breast cancer risk

Dooman Arefan et al. Med Phys. 2020 Jan.

Abstract

Purpose: To investigate two deep learning-based modeling schemes for predicting short-term risk of developing breast cancer using prior normal screening digital mammograms in a case-control setting.

Methods: We conducted a retrospective Institutional Review Board-approved study on a case-control cohort of 226 patients (including 113 women diagnosed with breast cancer and 113 controls) who underwent general population breast cancer screening. For each patient, a prior normal (i.e., with negative or benign findings) digital mammogram examination [including mediolateral oblique (MLO) view and craniocaudal (CC) view two images] was collected. Thus, a total of 452 normal images (226 MLO view images and 226 CC view images) of this case-control cohort were analyzed to predict the outcome, i.e., developing breast cancer (cancer cases) or remaining breast cancer-free (controls) within the follow-up period. We implemented an end-to-end deep learning model and a GoogLeNet-LDA model and compared their effects in several experimental settings using two mammographic view images and inputting two different subregions of the images to the models. The proposed models were also compared to logistic regression modeling of mammographic breast density. Area under the receiver operating characteristic curve (AUC) was used as the model performance metric.

Results: The highest AUC was 0.73 [95% Confidence Interval (CI): 0.68-0.78; GoogLeNet-LDA model on CC view] when using the whole-breast and was 0.72 (95% CI: 0.67-0.76; GoogLeNet-LDA model on MLO + CC view) when using the dense tissue, respectively, as the model input. The GoogleNet-LDA model significantly (all P < 0.05) outperformed the end-to-end GoogLeNet model in all experiments. CC view was consistently more predictive than MLO view in both deep learning models, regardless of the input subregions. Both models exhibited superior performance than the percent breast density (AUC = 0.54; 95% CI: 0.49-0.59).

Conclusions: The proposed deep learning modeling approach can predict short-term breast cancer risk using normal screening mammogram images. Larger studies are needed to further reveal the promise of deep learning in enhancing imaging-based breast cancer risk assessment.

Keywords: breast cancer; breast density; deep learning; digital mammography; risk biomarkers.

PubMed Disclaimer

Figures

Figure 1.
Figure 1.
The proposed schemes for deep learning-based modeling for short-term breast cancer risk prediction. The top half is an end-to-end prediction model using fine-tuned GoogLeNet, which is also adapted as an offline deep imaging feature extractor in the GoogLeNet-LDA model (bottom half).
Figure 2.
Figure 2.
Two different kinds of sub-regional inputs for the two deep learning models: One is the whole-breast region (red contours) and the other is the dense breast tissue only (green contours). The two regions were segmented using an automated computer method.
Figure 3.
Figure 3.
Six representative ROC curves for predicting short-term breast cancer risk: four from all the four experiments using CC view and two from using the MLO+CC view and dense tissue as input.
Figure 4.
Figure 4.
Feature map visualization for four selected sample images from the control (left two samples) and case (right two samples) groups. The color bar indicates the importance level (highest 100 and lowest 0) of a specific region in the images in predicting the short-term risk.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. D’orsi C, Bassett L, Feig S, Breast imaging reporting and data system (BI-RADS), Breast imaging atlas, 4th edn. American College of Radiology, Reston: (1998).
    1. Keller BM, Nathan DL, Wang Y, Zheng Y, Gee JC, Conant EF, Kontos D, Estimation of breast percent density in raw and processed full field digital mammography images via adaptive fuzzy c‐means clustering and support vector machine segmentation, Medical physics 39(8) (2012) 4903–4917. - PMC - PubMed
    1. Ciatto S, Bernardi D, Calabrese M, Durando M, Gentilini MA, Mariscotti G, Monetti F, Moriconi E, Pesce B, Roselli A, A first evaluation of breast radiological density assessment by QUANTRA software as compared to visual classification, The Breast 21(4) (2012) 503–506. - PubMed
    1. Alain G, Bengio Y, Understanding intermediate layers using linear classifier probes, arXiv preprint arXiv:1610.01644 (2016).
    1. Boyd NF, Guo H, Martin LJ, Sun L, Stone J, Fishell E, Jong RA, Hislop G, Chiarelli A, Minkin S, Mammographic density and the risk and detection of breast cancer, New England Journal of Medicine 356(3) (2007) 227–236. - PubMed