Mammographic density assessed on paired raw and processed digital images and on paired screen-film and digital images across three mammography systems

Association and Prediction Utilizing Craniocaudal and Mediolateral Oblique View Digital Mammography and Long-Term Breast Cancer Risk

Mammographic percentage of volumetric density is an important risk factor for breast cancer. Epidemiology studies historically used film images often limited to craniocaudal (CC) views to estimate area-based breast density. More recent studies using digital mammography images typically use the averaged density between craniocaudal (CC) and mediolateral oblique (MLO) view mammography for 5- and 10-year risk prediction. The performance in using either and both mammogram views has not been well-investigated. We use 3,804 full-field digital mammograms from the Joanne Knight Breast Health Cohort (294 incident cases and 657 controls), to quantity the association between volumetric percentage of density extracted from either and both mammography views and to assess the 5 and 10-year breast cancer risk prediction performance. Our results show that the association between percent volumetric density from CC, MLO, and the average between the two, retain essentially the same association with breast cancer risk. The 5- and 10-year risk prediction also shows similar prediction accuracy. Thus, one view is sufficient to assess association and predict future risk of breast cancer over a 5 or 10-year interval.

PREVENTION RELEVANCE

Expanding use of digital mammography and repeated screening provides opportunities for risk assessment. To use these images for risk estimates and guide risk management in real time requires efficient processing. Evaluating the contribution of different views to prediction performance can guide future applications for risk management in routine care.

Longitudinal Analysis of Change in Mammographic Density in Each Breast and Its Association With Breast Cancer Risk

Importance

Although breast density is an established risk factor for breast cancer, longitudinal changes in breast density have not been extensively studied to determine whether this factor is associated with breast cancer risk.

Objective

To prospectively evaluate the association between change in mammographic density in each breast over time and risk of subsequent breast cancer.

Design, Setting, and Participants

This nested case-control cohort study was sampled from the Joanne Knight Breast Health Cohort of 10 481 women free from cancer at entry and observed from November 3, 2008, to October 31, 2020, with routine screening mammograms every 1 to 2 years, providing a measure of breast density. Breast cancer screening was provided for a diverse population of women in the St Louis region. A total of 289 case patients with pathology-confirmed breast cancer were identified, and approximately 2 control participants were sampled for each case according to age at entry and year of enrollment, yielding 658 controls with a total number of 8710 craniocaudal-view mammograms for analysis.

Exposures

Exposures included screening mammograms with volumetric percentage of density, change in volumetric breast density over time, and breast biopsy pathology-confirmed cancer. Breast cancer risk factors were collected via questionnaire at enrollment.

Main Outcomes and Measures

Longitudinal changes over time in each woman's volumetric breast density by case and control status.

Results

The mean (SD) age of the 947 participants was 56.67 (8.71) years at entry; 141 were Black (14.9%), 763 were White (80.6%), 20 were of other race or ethnicity (2.1%), and 23 did not report this information (2.4%). The mean (SD) interval was 2.0 (1.5) years from last mammogram to date of subsequent breast cancer diagnosis (10th percentile, 1.0 year; 90th percentile, 3.9 years). Breast density decreased over time in both cases and controls. However, there was a significantly slower decrease in rate of decline in density in the breast that developed breast cancer compared with the decline in controls (estimate = 0.027; 95% CI, 0.001-0.053; P = .04).

Conclusions and Relevance

This study found that the rate of change in breast density was associated with the risk of subsequent breast cancer. Incorporation of longitudinal changes into existing models could optimize risk stratification and guide more personalized risk management.

The association of age at menarche and adult height with mammographic density in the International Consortium of Mammographic Density

BACKGROUND

Early age at menarche and tall stature are associated with increased breast cancer risk. We examined whether these associations were also positively associated with mammographic density, a strong marker of breast cancer risk.

METHODS

Participants were 10,681 breast-cancer-free women from 22 countries in the International Consortium of Mammographic Density, each with centrally assessed mammographic density and a common set of epidemiologic data. Study periods for the 27 studies ranged from 1987 to 2014. Multi-level linear regression models estimated changes in square-root per cent density (√PD) and dense area (√DA) associated with age at menarche and adult height in pooled analyses and population-specific meta-analyses. Models were adjusted for age at mammogram, body mass index, menopausal status, hormone therapy use, mammography view and type, mammographic density assessor, parity and height/age at menarche.

RESULTS

In pooled analyses, later age at menarche was associated with higher per cent density (β√PD = 0.023 SE = 0.008, P = 0.003) and larger dense area (β√DA = 0.032 SE = 0.010, P = 0.002). Taller women had larger dense area (β√DA = 0.069 SE = 0.028, P = 0.012) and higher per cent density (β√PD = 0.044, SE = 0.023, P = 0.054), although the observed effect on per cent density depended upon the adjustment used for body size. Similar overall effect estimates were observed in meta-analyses across population groups.

CONCLUSIONS

In one of the largest international studies to date, later age at menarche was positively associated with mammographic density. This is in contrast to its association with breast cancer risk, providing little evidence of mediation. Increased height was also positively associated with mammographic density, particularly dense area. These results suggest a complex relationship between growth and development, mammographic density and breast cancer risk. Future studies should evaluate the potential mediation of the breast cancer effects of taller stature through absolute breast density.

Correlation between cumulative mammographic density and age-specific incidence of breast cancer: A biethnic study in Israel

Women with the most extensive breast density, have a 4- to 6-fold higher cancer risk than women with the lowest density. This cross-sectional study evaluated associations of cumulative mammographic density in two distinct ethnic groups with the respective age-specific breast cancer incidences in the population. The study compared four cohorts of 200 women each aged 35 to 49 and 50 to 74, representing Jewish and Arab ethnicity. Breast density measures were calculated from screening mammograms, using a thresholding software (Cumulus). Breast cancer specific incidence values were obtained from the National Cancer Registry. The percent mammographic density was lower for women aged 50 to 74 than 35 to 49 years, both for Jews: 11.7 vs 23.1 and for Arabs: 11.6 vs 18.3. In contrast, the cumulative density increased with age, from 37.30 to 181.24 in Jews, compared to 21.26 to 108.03 in Arabs. Similar trends in breast cancer incidence rates per 100 000 in the Israeli population were apparent, with an increase from 92.95 to 381.91 in Jews, compared to 48.6 to 244.44 in Arabs. Comparing cumulative density of the cohort with respective age-specific breast cancer incidence in the population yielded a highly significant correlation: Jews; r = .97, P < .0001 and Arabs: r = .86, P = .007. A strong association was found between the log of cumulative density and the log of cancer incidence, as well. Our study identified correlations between cumulative mammographic density and breast cancer incidence in two distinct populations. The findings should prompt research to enhance our understanding of the pathogenesis of breast cancer, and lead to novel insights into measures of prevention.

Deep learning identification of stiffness markers in breast cancer

While essential to our understanding of solid tumor progression, the study of cell and tissue mechanics has yet to find traction in the clinic. Determining tissue stiffness, a mechanical property known to promote a malignant phenotype in vitro and in vivo, is not part of the standard algorithm for the diagnosis and treatment of breast cancer. Instead, clinicians routinely use mammograms to identify malignant lesions and radiographically dense breast tissue is associated with an increased risk of developing cancer. Whether breast density is related to tumor tissue stiffness, and what cellular and non-cellular components of the tumor contribute the most to its stiffness are not well understood. Through training of a deep learning network and mechanical measurements of fresh patient tissue, we create a bridge in understanding between clinical and mechanical markers. The automatic identification of cellular and extracellular features from hematoxylin and eosin (H&E)-stained slides reveals that global and local breast tissue stiffness best correlate with the percentage of straight collagen. Importantly, the percentage of dense breast tissue does not directly correlate with tissue stiffness or straight collagen content.

Mammographic texture features associated with contralateral breast cancer in the WECARE Study

To evaluate whether mammographic texture features were associated with second primary contralateral breast cancer (CBC) risk, we created a "texture risk score" using pre-treatment mammograms in a case-control study of 212 women with CBC and 223 controls with unilateral breast cancer. The texture risk score was associated with CBC (odds per adjusted standard deviation = 1.25, 95% CI 1.01-1.56) after adjustment for mammographic percent density and confounders. These results support the potential of texture features for CBC risk assessment of breast cancer survivors.

A Deep Learning Approach to Re-create Raw Full-Field Digital Mammograms for Breast Density and Texture Analysis

Purpose

To develop a computational approach to re-create rarely stored for-processing (raw) digital mammograms from routinely stored for-presentation (processed) mammograms.

Materials and Methods

In this retrospective study, pairs of raw and processed mammograms collected in 884 women (mean age, 57 years ± 10 [standard deviation]; 3713 mammograms) from October 5, 2017, to August 1, 2018, were examined. Mammograms were split 3088 for training and 625 for testing. A deep learning approach based on a U-Net convolutional network and kernel regression was developed to estimate the raw images. The estimated raw images were compared with the originals by four image error and similarity metrics, breast density calculations, and 29 widely used texture features.

Results

In the testing dataset, the estimated raw images had small normalized mean absolute error (0.022 ± 0.015), scaled mean absolute error (0.134 ± 0.078) and mean absolute percentage error (0.115 ± 0.059), and a high structural similarity index (0.986 ± 0.007) for the breast portion compared with the original raw images. The estimated and original raw images had a strong correlation in breast density percentage (Pearson r = 0.946) and a strong agreement in breast density grade (Cohen κ = 0.875). The estimated images had satisfactory correlations with the originals in 23 texture features (Pearson r ≥ 0.503 or Spearman ρ ≥ 0.705) and were well complemented by processed images for the other six features.

Conclusion

This deep learning approach performed well in re-creating raw mammograms with strong agreement in four image evaluation metrics, breast density, and the majority of 29 widely used texture features.Keywords: Mammography, Breast, Supervised Learning, Convolutional Neural Network (CNN), Deep learning algorithms, Machine Learning AlgorithmsSee also the commentary by Chan in this issue.Supplemental material is available for this article.©RSNA, 2021.

The association between mammographic density and breast cancer risk in Western Australian Aboriginal women

PURPOSE

Mammographic density is an established breast cancer risk factor within many ethnically different populations. The distribution of mammographic density has been shown to be significantly lower in Western Australian Aboriginal women compared to age- and screening location-matched non-Aboriginal women. Whether mammographic density is a predictor of breast cancer risk in Aboriginal women is unknown.

METHODS

We measured mammographic density from 103 Aboriginal breast cancer cases and 327 Aboriginal controls, 341 non-Aboriginal cases, and 333 non-Aboriginal controls selected from the BreastScreen Western Australia database using the Cumulus software program. Logistic regression was used to examine the associations of percentage dense area and absolute dense area with breast cancer risk for Aboriginal and non-Aboriginal women separately, adjusting for covariates.

RESULTS

Both percentage density and absolute dense area were strongly predictive of risk in Aboriginal women with odds per adjusted standard deviation (OPERAS) of 1.36 (95% CI 1.09, 1.69) and 1.36 (95% CI 1.08, 1.71), respectively. For non-Aboriginal women, the OPERAS were 1.22 (95% CI 1.03, 1.46) and 1.26 (95% CI 1.05, 1.50), respectively.

CONCLUSIONS

Whilst mean mammographic density for Aboriginal women is lower than non-Aboriginal women, density measures are still higher in Aboriginal women with breast cancer compared to Aboriginal women without breast cancer. Thus, mammographic density strongly predicts breast cancer risk in Aboriginal women. Future efforts to predict breast cancer risk using mammographic density or standardize risk-associated mammographic density measures should take into account Aboriginal status when applicable.

The distribution and determinants of mammographic density measures in Western Australian aboriginal women

Background

Mammographic density (MD) is an established risk factor for breast cancer. There are significant ethnic differences in MD measures which are consistent with those for corresponding breast cancer risk. This is the first study investigating the distribution and determinants of MD measures within Aboriginal women of Western Australia (WA).

Methods

Epidemiological data and mammographic images were obtained from 628 Aboriginal women and 624 age-, year of screen-, and screening location-matched non-Aboriginal women randomly selected from the BreastScreen Western Australia database. Women were cancer free at the time of their mammogram between 1989 and 2014. MD was measured using the Cumulus software. Kolmogorov-Smirnov tests were used to compare distributions of absolute dense area (DA), precent dense area (PDA), non-dense area (NDA) and total breast area between Aboriginal and non-Aboriginal women. General linear regression was used to estimate the determinants of MD, adjusting for age, NDA, hormone therapy use, family history, measures of socio-economic status and remoteness of residence for Aboriginal and non-Aboriginal women separately.

Results

Aboriginal women were found to have lower DA and PDA and higher NDA than non-Aboriginal women. Age (p <  0.001) was negatively associated and several socio-economic indices (p <  0.001) were positively associated with DA and PDA in Aboriginal and non-Aboriginal women. Remoteness of residence was associated with both mammographic measures but for non-Aboriginal women only.

Conclusions

Aboriginal women have, on average, less MD than non-Aboriginal women but the factors associated with MD are similar for both sample populations. Since reduced MD is associated with improved sensitivity of mammography, this study suggests that mammographic screening is a particularly good test for Australian Indigenous women, a population that suffers from high breast cancer mortality.

Electronic supplementary material

The online version of this article (10.1186/s13058-019-1113-4) contains supplementary material, which is available to authorized users.

Automated mammographic breast density estimation using a fully convolutional network

PURPOSE

The purpose of this study was to develop a fully automated algorithm for mammographic breast density estimation using deep learning.

METHOD

Our algorithm used a fully convolutional network, which is a deep learning framework for image segmentation, to segment both the breast and the dense fibroglandular areas on mammographic images. Using the segmented breast and dense areas, our algorithm computed the breast percent density (PD), which is the faction of dense area in a breast. Our dataset included full-field digital screening mammograms of 604 women, which included 1208 mediolateral oblique (MLO) and 1208 craniocaudal (CC) views. We allocated 455, 58, and 91 of 604 women and their exams into training, testing, and validation datasets, respectively. We established ground truth for the breast and the dense fibroglandular areas via manual segmentation and segmentation using a simple thresholding based on BI-RADS density assessments by radiologists, respectively. Using the mammograms and ground truth, we fine-tuned a pretrained deep learning network to train the network to segment both the breast and the fibroglandular areas. Using the validation dataset, we evaluated the performance of the proposed algorithm against radiologists' BI-RADS density assessments. Specifically, we conducted a correlation analysis between a BI-RADS density assessment of a given breast and its corresponding PD estimate by the proposed algorithm. In addition, we evaluated our algorithm in terms of its ability to classify the BI-RADS density using PD estimates, and its ability to provide consistent PD estimates for the left and the right breast and the MLO and CC views of the same women. To show the effectiveness of our algorithm, we compared the performance of our algorithm against a state of the art algorithm, laboratory for individualized breast radiodensity assessment (LIBRA).

RESULT

The PD estimated by our algorithm correlated well with BI-RADS density ratings by radiologists. Pearson's rho values of our algorithm for CC view, MLO view, and CC-MLO-averaged were 0.81, 0.79, and 0.85, respectively, while those of LIBRA were 0.58, 0.71, and 0.69, respectively. For CC view and CC-MLO averaged cases, the difference in rho values between the proposed algorithm and LIBRA showed statistical significance (P < 0.006). In addition, our algorithm provided reliable PD estimates for the left and the right breast (Pearson's ρ > 0.87) and for the MLO and CC views (Pearson's ρ = 0.76). However, LIBRA showed a lower Pearson's rho value (0.66) for both the left and right breasts for the CC view. In addition, our algorithm showed an excellent ability to separate each sub BI-RADS breast density class (statistically significant, p-values = 0.0001 or less); only one comparison pair, density 1 and density 2 in the CC view, was not statistically significant (P = 0.54). However, LIBRA failed to separate breasts in density 1 and 2 for both the CC and MLO views (P > 0.64).

CONCLUSION

We have developed a new deep learning based algorithm for breast density segmentation and estimation. We showed that the proposed algorithm correlated well with BI-RADS density assessments by radiologists and outperformed an existing state of the art algorithm.

A comprehensive tool for measuring mammographic density changes over time

Background

Mammographic density is a marker of breast cancer risk and diagnostics accuracy. Density change over time is a strong proxy for response to endocrine treatment and potentially a stronger predictor of breast cancer incidence. We developed STRATUS to analyse digital and analogue images and enable automated measurements of density changes over time.

Method

Raw and processed images from the same mammogram were randomly sampled from 41,353 healthy women. Measurements from raw images (using FDA approved software iCAD) were used as templates for STRATUS to measure density on processed images through machine learning. A similar two-step design was used to train density measures in analogue images. Relative risks of breast cancer were estimated in three unique datasets. An alignment protocol was developed using images from 11,409 women to reduce non-biological variability in density change. The protocol was evaluated in 55,073 women having two regular mammography screens. Differences and variances in densities were compared before and after image alignment.

Results

The average relative risk of breast cancer in the three datasets was 1.6 [95% confidence interval (CI) 1.3–1.8] per standard deviation of percent mammographic density. The discrimination was AUC 0.62 (CI 0.60–0.64). The type of image did not significantly influence the risk associations. Alignment decreased the non-biological variability in density change and re-estimated the yearly overall percent density decrease from 1.5 to 0.9%, p < 0.001.

Conclusions

The quality of STRATUS density measures was not influenced by mammogram type. The alignment protocol reduced the non-biological variability between images over time. STRATUS has the potential to become a useful tool for epidemiological studies and clinical follow-up.

Electronic supplementary material

The online version of this article (10.1007/s10549-018-4690-5) contains supplementary material, which is available to authorized users.

Mammographic density and ageing: A collaborative pooled analysis of cross-sectional data from 22 countries worldwide

BACKGROUND

Mammographic density (MD) is one of the strongest breast cancer risk factors. Its age-related characteristics have been studied in women in western countries, but whether these associations apply to women worldwide is not known.

METHODS AND FINDINGS

We examined cross-sectional differences in MD by age and menopausal status in over 11,000 breast-cancer-free women aged 35-85 years, from 40 ethnicity- and location-specific population groups across 22 countries in the International Consortium on Mammographic Density (ICMD). MD was read centrally using a quantitative method (Cumulus) and its square-root metrics were analysed using meta-analysis of group-level estimates and linear regression models of pooled data, adjusted for body mass index, reproductive factors, mammogram view, image type, and reader. In all, 4,534 women were premenopausal, and 6,481 postmenopausal, at the time of mammography. A large age-adjusted difference in percent MD (PD) between post- and premenopausal women was apparent (-0.46 cm [95% CI: -0.53, -0.39]) and appeared greater in women with lower breast cancer risk profiles; variation across population groups due to heterogeneity (I2) was 16.5%. Among premenopausal women, the √PD difference per 10-year increase in age was -0.24 cm (95% CI: -0.34, -0.14; I2 = 30%), reflecting a compositional change (lower dense area and higher non-dense area, with no difference in breast area). In postmenopausal women, the corresponding difference in √PD (-0.38 cm [95% CI: -0.44, -0.33]; I2 = 30%) was additionally driven by increasing breast area. The study is limited by different mammography systems and its cross-sectional rather than longitudinal nature.

CONCLUSIONS

Declines in MD with increasing age are present premenopausally, continue postmenopausally, and are most pronounced over the menopausal transition. These effects were highly consistent across diverse groups of women worldwide, suggesting that they result from an intrinsic biological, likely hormonal, mechanism common to women. If cumulative breast density is a key determinant of breast cancer risk, younger ages may be the more critical periods for lifestyle modifications aimed at breast density and breast cancer risk reduction.

Qualitative Versus Quantitative Mammographic Breast Density Assessment: Applications for the US and Abroad

Mammographic breast density (MBD) has been proven to be an important risk factor for breast cancer and an important determinant of mammographic screening performance. The measurement of density has changed dramatically since its inception. Initial qualitative measurement methods have been found to have limited consistency between readers, and in regards to breast cancer risk. Following the introduction of full-field digital mammography, more sophisticated measurement methodology is now possible. Automated computer-based density measurements can provide consistent, reproducible, and objective results. In this review paper, we describe various methods currently available to assess MBD, and provide a discussion on the clinical utility of such methods for breast cancer screening.

Percent mammographic density prediction: development of a model in the nurses' health studies

PURPOSE

To develop a model to predict percent mammographic density (MD) using questionnaire data and mammograms from controls in the Nurses' Health Studies' nested breast cancer case-control studies. Further, we assessed the association between both measured and predicted percent MD and breast cancer risk.

METHODS

Using data from 2,955 controls, we assessed several variables as potential predictors. We randomly divided our dataset into a training dataset (two-thirds of the dataset) and a testing dataset (one-third of the dataset). We used stepwise linear regression to identify the subset of variables that were most predictive. Next, we examined the correlation between measured and predicted percent MD in the testing dataset and computed the r ² in the total dataset. We used logistic regression to examine the association between measured and predicted percent MD and breast cancer risk.

RESULTS

In the training dataset, several variables were selected for inclusion, including age, body mass index, and parity, among others. In the testing dataset, the Spearman correlation coefficient between predicted and measured percent MD was 0.61. As the prediction model performed well in the testing dataset, we developed the final model in the total dataset. The final prediction model explained 41% of the variability in percent MD. Both measured and predicted percent MD were similarly associated with breast cancer risk adjusting for age, menopausal status, and hormone use (OR per five unit increase = 1.09 for both).

CONCLUSION

These results suggest that predicted percent MD may be useful for research studies in which mammograms are unavailable.

Visualization of Breast Microcalcifications on Digital Breast Tomosynthesis and 2-Dimensional Digital Mammography Using Specimens

Purpose:

The purpose of this study is to compare the visibility of microcalcifications of digital breast tomosynthesis (DBT) and full-field digital mammography (FFDM) using breast specimens.

Materials And Methods:

Thirty-one specimens’ DBT and FFDM were retrospectively reviewed by four readers.

Results:

The image quality of microcalcifications of DBT was rated as superior or equivalent in 71.0% by reader 1, 67.8% by reader 2, 64.5% by reader 3, and 80.6% by reader 4. The Fleiss kappa statistic for agreement among readers was 0.31.

Conclusions:

We suggest that image quality of DBT appears to be comparable with or better than FFDM in terms of revealing microcalcifications.