Full-field Digital Mammograms Research Articles

Architectural distortion on digital breast tomosynthesis mammograms in symptomatic breast clinics: What are the result outcomes?

In April 2020, standard two-dimensional Full-field Digital Mammograms (FFDM) were replaced with Digital Breast Tomosynthesis (DBT) and synthesised 2D views for symptomatic breast clinics. This study aimed to evaluate the positive predictive value for malignancy in DBT-detected Architectural Distortion (AD). All mammogram reports with the word 'distortion' were assessed between April 2020 and October 2022. There were 458 mammograms with the word 'distortion'. After excluding mammograms with no distortion (n = 128), post-surgical distortion (n = 173), distortion with mass (n = 33), and unchanged distortion (n = 14), there were 111 patients with pure distortion. Correlation with histopathology was obtained where possible. All patients were followed for a minimum of two years. Forty-two out of 111 patients (37.84%) with AD had a normal ultrasound(US) and were discharged. Fifty-five (49.5%) patients had sonographic correlation corresponding to the distortion, leading to US-guided biopsy. Thirteen (23.6%) had Tomosynthesis-guided biopsy, and one had a skin biopsy. The positive predictive value (PPV) for malignancy was 42.34%. Malignancy diagnoses were higher with US-guided biopsies than Tomosynthesis-guided biopsies, 78.1% and 30%, respectively. With a total malignancy rate of 42.34%, DBT-detected Architectural Distortion has a high enough PPV for malignancy to justify selective tissue sampling if a sonographic correlate is present or with suspicious mammograms. The chances of malignancy are higher when a sonographic correlate corresponding to AD is present. Architectural Distortion on DBT/SM views in Symptomatic Breast clinic patients justifies selective sampling.

Abstract PS10-08: Validation of a clinical image-based AI-risk model for individualized breast cancer screening in a multi-national setting

Abstract Background: Image-derived AI risk models for breast cancer have shown high discriminatory performances compared with clinical risk models based on family history and lifestyle factors. However, little is known about their generalizability across different screening settings and clinical feasibility. Purpose: To investigate the predictive performance of a clinically used image-derived AI-based breast cancer risk model in multiple European screening populations. Methods: Four European mammographic screening populations in three countries screened between 2009-2020 for women aged 45-69 was used to perform a nested case-control study. In total, 739 women with incident breast cancers were included together with 7,812 controls matched to cases on year of study-entry. Mammographic features (density, microcalcifications, masses, left-right breast asymmetries of these features) for risk assessment were extracted using AI from full-field digital mammograms. Breast cancer occurrence was assessed after two years of follow-up. Absolute risks of breast cancer were predicted using the risk model from negative mammograms at study-entry. Adjusted Area Under the receiver operating characteristic Curves (aAUC) estimated discriminatory performance and, adjusted risk-ratios estimated the stratification performance of women at high/general risk per the clinical guidelines. Results: The overall aAUC of the AI risk model was 0.72 (95%CI 0.70-0.75), range 0.71 (95%CI 0.67-0.75) to 0.74 (95%CI 0.69-0.78) for breast cancers developed in four screening populations. In the 4.6% of women classified at high risk using the NICE guidelines thresholds, cancers were more likely diagnosed after 2 years follow-up, risk-ratio (RR) 6.7 (95%CI 5.6-8.0), compared with the 71% of women classified at general risk by the model. Similar risk-ratios were observed across tertiles of mammographic density. In the high-risk group, 22% of the 2-year future cancers were diagnosed, and 29% of stage 2 and higher cancers, p< 0.01. Conclusions: The AI risk model showed generalizable discriminatory performances across European populations and, captured ~30% of clinically relevant stage 2 and higher breast cancers in ~5% of high-risk women who were sent home with a negative mammogram. Similar results were seen in fatty and dense breasts. An image-derived AI model is feasible for personalized screening to improve screening outcomes. Citation Format: Mikael Eriksson, Marta Román, Axel Gräwingholt, Sylvia Heywang-Köbrunner, Paolo Rossi, Per Hall. Validation of a clinical image-based AI-risk model for individualized breast cancer screening in a multi-national setting [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PS10-08.

Abstract PO3-09-01: Integrating pathomic and radiomic images to classify risk of subsequent events among women with DCIS

Abstract Background Multiple sources of ~omic data can be generated from women at different stages of developing breast cancer, the leading cancer diagnosed in women worldwide. Traditionally interrogation of risk factors to study associations and develop prediction models for future breast events has been limited to one or few risk factors, or summary scores of clinical and tumor characteristics. Methods to bring mammography images and breast biopsies of precancer lesions together to summarize risk of cancer developing in the breast are urgently needed. Integration of these two sources has not been performed to date, but has potential to increase accuracy of risk prediction. Approach The Repository of Archival Human Breast Tissue (RAHBT) was established in 2007 at Washington University School of Medicine (WUSM) and maintains biospecimens and medical record data of women treated with breast-conserving surgery (BCS) or mastectomy for breast cancer at WUSM and six other St Louis metropolitan hospitals between 1981-2016. Prospective follow-up of cases is achieved through health records and Siteman tumor registry. Among 1831 patients with pathologically confirmed DCIS who had no prior cancer, 174 developed breast events at least six months after initial DCIS diagnosis. For each case diagnosed between January 1998 and March 2016 with subsequent breast events, two DCIS controls were matched on race, year of diagnosis (±5 years), age (±5 years), and type of surgery. Tissue micro arrays (TMAs) are constructed after breast pathology review and processed for H&E and imaged. Breast cancer risk factor data are uniformly retrieved form the medical records at time of processing cases of DCIS. Mammogram images are retrieved for all women. Full-field digital mammograms (FFDM) are all using the same technology (Hologic). Risk factors and outcome variables: ipsilateral breast events (IBE) was defined as the development of invasive cancer or DCIS in the treated breast at least 6 months post-diagnosis. Follow-up time was calculated from the date of DCIS diagnosis to the date of first IBE, death, or last follow-up, whichever occurred first. The predictors included commonly used clinicopathological factors: age, tumor grade (high vs. low/intermediate), comedonecrosis, surgical margins (close (≤2mm) vs. negative ( >2mm)), local treatment (BCS only and mastectomy vs. BCS+radiation), and endocrine therapy. BMI, menopausal status, and ER were also available for evaluation as predictors. We limited this analysis to women with digital mammograms immediately prior to diagnosis and include 128 cases and controls. Validation cohort As the RAHBT cohort continues to be followed additional second breast events have been documented after the cut off for events identified through March 2016, and the identical procedures used to review/confirm DCIS and subsequent breast events, construct TMAs, process H&E and images, and assemble risk factor data, and the associated digital mammograms. We have 13 additional ipsilateral breast events with complete data and their controls for independently testing the model performance. Statistical methods Propose novel supervised learning approach to integrate image data from mammograms and TMA slides that accommodates agreement between multi-omics. Results 128 (cases and controls, 21.9% ipsilateral subsequent events) identified for development and internal cross validation. Median age at diagnosis was 52 and median time to subsequent breast event was 63 mo. Controlling for the clinical/pathologic risk factors (age, BMI, BI-RADs density, treatment, tumor grade, parous, menopausal status, ER status, and race) and integrating the whole mammogram and TMA pathology images, we observe a 5-yr ipsilateral AUC = 0.74; 10-yr AUC 0.75. The independent validation is ongoing. Conclusions Integrating heterogeneous multiomic data sources can generate significant improvement in long term risk prediction after initial DCIS diagnosis. Citation Format: Shih-Ting Huang, Debbie Bennett, Robert West, Graham Colditz, Shu Jiang. Integrating pathomic and radiomic images to classify risk of subsequent events among women with DCIS [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO3-09-01.

Abstract PO2-09-04: Changes in an artificial intelligence model score for breast cancer detection (Transpara) with tamoxifen and aromatase inhibitor treatment

Abstract Artificial intelligence (AI) models applied to mammography have been shown to improve breast cancer (BC) detection and risk estimation. The Transpara AI detection system has been associated with short and longer-term risk of invasive BC. Whether Transpara can identify changes in the breast associated with tamoxifen (TAM) and aromatase inhibitor (AIs) treatment is unknown. We compare changes in the Transpara malignancy score in the unaffected breast of BC cases using TAM or AIs to those of untreated women without cancer. Cases and controls were sampled from the Mayo Clinic Rochester Breast Screening Practice. Eligible cases were women with breast cancer treated with TAM and/or AIs for at least 8 months who had full field digital (FFDM) mammograms within two years prior to start of treatment (pretreatment mammogram). Controls did not have BC, were not using TAM, AIs, or postmenopausal hormones, and had at least two FFDM screening mammograms. Transpara scores (1-10) were calculated for the contralateral breast in cases and a randomly selected side for controls. Two controls were matched to each case on menopausal status at both mammograms, time between mammograms (< 6 months) and baseline Transpara score (±2 units). Conditional logistic regression was used to assess the association of treatment (case (TAM/AIs) vs. control) with change in Transpara score between the pretreatment mammogram and the mammogram taken closest to 13 months later. Change was classified as having a decrease in Transpara score between the two mammograms vs. no change or an increase. Models were adjusted for age and Transpara score at pretreatment mammogram, and time between the two mammograms. We performed analyses on all women combined and stratified by time on treatment, menopausal status, and treatment type (TAM vs. AIs). A total of 134 cases were identified, with 53 on TAM (21 premenopausal and 32 postmenopausal) and 81 on AIs (postmenopausal); 268 controls were matched to the cases. Characteristics of the study population are summarized (Table). Most case mammograms were obtained within six months prior to treatment start [median 2.63 months (interquartile range (IQR), 1.62, 5.72)] and median time on TAM/AIs was 10.9 months (IQR, 8.03, 16.6) at time of second mammogram. The median absolute change in Transpara score between the two mammograms was similar in cases [0 (IQR, -2, 1)] vs. controls [0 (IQR, -1.25, 1)]; 40.3% of cases and 39.2% of controls had any decrease. Conditional logistic models found no evidence for an association between TAM/AIs and decrease in Transpara score [odds ratio (OR) = 1.07; 95% confidence interval (CI), 0.65, 1.78]. When stratified by median time on treatment (10.9 months), there was suggestion of a stronger association for those on treatment for a longer duration (OR = 1.21; 95% CI, 0.61, 2.40 for >10.9 months vs. OR = 0.96; 95% CI, 0.43, 2.13 for < 10.9 months) but these results were not statistically significant. Similarly, odds ratios were stronger in postmenopausal (OR = 1.18; 95% CI, 0.67, 2.07) compared to premenopausal (OR = 0.91; 95% CI, 0.24, 3.45) women, and for AIs (OR = 1.21, 95% CI, 0.64, 2.29) vs. TAM (OR = 0.93; 95% CI, 0.39, 2.22) but none of these results reached statistical significance. In summary, we found no association between treatment with TAM/AIs and change in Transpara score. However, our study had limited power. Also, longer time on treatment as well as investigation of AI-based scores trained directly to assess BC risk may be necessary to identify relevant changes. Citation Format: Yvonne Wang, Christopher Scott, Matthew Jensen, Dan Hursh, Kathleen Brandt, Aaron Norman, Fang Fang Wu, Stacey Winham, Karla Kerlikowske, Celine Vachon. Changes in an artificial intelligence model score for breast cancer detection (Transpara) with tamoxifen and aromatase inhibitor treatment [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO2-09-04.

Abstract PS10-06: Contribution of a breast cancer polygenic risk score to mammography artificial intelligence models and breast density for long term breast cancer risk prediction

Abstract Artificial intelligence (AI) models based on deep learning of mammogram images have been developed for lesion detection and diagnosis as well as breast cancer (BC) risk. We and others have shown that these AI models, coupled with breast density, predict long-term breast cancer risk. Common germline genetic variation, in particular a polygenic risk score (PRS) that captures the cumulative impact of hundreds of common risk variants, has been shown associated with 3-fold increased BC risk. The BC-PRS may complement imaging risk factors to improve identification of women at high BC risk and better target prevention strategies. We evaluated the contribution of a BC-PRS to two imaging-based AI models (Transpara AI cancer detection system and Mirai 5-year risk) and breast density measures to long-term BC risk. We hypothesized that the BC-PRS will improve long-term risk prediction above AI and density measures. We examined this hypothesis within a cohort of 10,271 women enrolled to the Mayo Clinic Biobank between 2009 and 2015. Eligible women had no prior cancer, were ages 35 to 85, had a full-field digital screening mammogram within two years of enrollment and genotyping by Regeneron. Incident invasive and DCIS BC was identified through the Mayo tumor registry and follow-up questionnaires, and person years were calculated from date of mammogram through date of BC, last mammogram at Mayo or last questionnaire. We estimated Transpara AI malignancy score (1-10), 5-year Mirai risk and Volpara volumetric density measures (volumetric percent density, dense volume) on screening mammograms at enrollment and obtained clinical BI-RADS density. We calculated the BC-PRS using 268 SNPs (of the 313-SNP score) and the weighted sum of the SNPs using odds ratios from the largest BC GWAS of European ancestry to date. We performed cox proportional hazards regression adjusting for age and BMI to estimate hazard ratios (HRs), 95% confidence intervals (CI), and C-statistics (AUC) to describe the contribution of BC-PRS (per standard deviation (SD)) to the association of AI scores and breast density with BC risk. Likelihood ratio tests (LRT) and bootstrapping were used to compare model performance with vs. without inclusion of BC-PRS, at 10-years follow-up. Comparisons to the Tyrer-Cuzick model are ongoing. Over a median of 9.1 years (Interquartile range, 6.8 to 10.6), 446 incident BC were identified of which 285 invasive and 103 DCIS were confirmed by pathology report (60 were self-reported); 399 (256 invasive) occurred within 10 years and were used in primary analyses. Transpara score [HR per 1 unit=1.18 (1.14, 1.23), AUC=0.655 (0.628, 0.682)] and Mirai risk [HR per SD log risk=1.54 (1.43, 1.65), AUC=0.681 (0.655, 0.706)] were both associated with BC. All three breast density measures were also significantly associated with BC, including in models with AI scores (Table). Addition of the BC-PRS to models with AI scores and breast density improved both model fit (All PLRT < 0.001; Table) and discrimination, with increases in AUC ranging from 0.033 to 0.034 for models with Transpara and density and 0.025 to 0.026 for models with Mirai and density (all P-values < 0.001; Table). Analyses restricted to invasive BC showed identical results. But, the BC-PRS contribution at 5-year follow-up was attenuated. In this cohort study, we found BC-PRS contributed to long-term risk prediction of BC above imaging AI models and breast density. Imaging and genetic risk factors are complementary and may outperform existing risk models and provide more accurate risk assessment to aid in precision prevention strategies. Citation Format: Celine Vachon, Christopher Scott, Imon Banerjee, Ramon Correa Moderno, Dan Hursh, Matthew Jensen, Yvonne Wang, Kathleen Brandt, Karla Kerlikowske, Sandhya Pruthi, Fergus Couch, Stacey Winham. Contribution of a breast cancer polygenic risk score to mammography artificial intelligence models and breast density for long term breast cancer risk prediction [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PS10-06.

Abstract PO2-23-10: Obesity May Modulate the Relationship Between High Mammographic Breast Density and Neoadjuvant Chemoresistance in Early Breast Cancer

Abstract Background - An inverse association has been described in a number of studies between mammographic breast density (MBD) and pathologic complete response (pCR) following neoadjuvant chemotherapy (NAC) for early breast cancer (BC), implicating high MBD in chemotherapy resistance. However, this effect has not been seen in all there is a paucity of data exploring whether this relationship extends the relationship beyond pCR to overall survival. This study aims to validate the relationship between MBD and pCR in patients undergoing NAC for early BC as well as to assess the relation of MBD with clinical complete response (cCR), relapse-free (RFS) and BC-specific survival (BCSS). Further, this study examines these associations stratified by body mass index (BMI). Methods - MBD was measured in contralateral mammograms in 127 women before NAC, using Cumulus software. As the study spanned the transition period from film to digital mammography, film mammograms were digitised using a high-powered scanner and processed full-field digital mammograms were extracted from the imaging storage facility. Basic patient, tumour and treatment demographics were collated including BMI. Percent dense area (PDA) was correlated with patient and tumour characteristics, short-term (pCR/cCR incidences), and long-term chemotherapy outcomes (RFS and BCSS). Results - Overall rates of cCR and pCR of 49% and 21% were observed across the whole cohort. Mean PDA was higher in those not undergoing cCR (p=0.041) and relapsing patients (p=0.041) but did not significantly vary by pCR or BC-deaths. As a dichotomous variable, high PDA relative to low PDA corresponded to lower cCR (40 v 58%, p=0.027), but no significant difference was seen for pCR (17.5 v 25.0%, p=0.15), BC relapse (38 v 30%, p=0.15) or BC-death (32 v 25%, p=0.20). High PDA (relative to low PDA) in patients with obesity (BMI >30kg/m2) was associated with lower pCR (0% v 28.1%, p=0.036) and numerically higher relapse rates (56% v 28%, p=0.063) and breast cancer deaths (56 v 28%, p=0.071. No such relationship was observed in patients with a BMI< 30kg/m2. Conclusion - Higher MBD associated with lower cCR and higher relapse but not with pCR or BC death. The inferior outcomes with high MBD were confined to patients with obesity in our study. Further analysis of the relationship between MBD and outcomes, including by BMI, is warranted. Citation Format: Andrew Redfern, Veenoo Agarwal, Ellie Darcey, Lisa Spalding, Hilary Martin, Jennifer Stone. Obesity May Modulate the Relationship Between High Mammographic Breast Density and Neoadjuvant Chemoresistance in Early Breast Cancer [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO2-23-10.

Abstract PO5-07-06: Impact of patient-specific factors on quantitative breast parenchymal texture features

Abstract Breast parenchymal texture characterizes the spatial and structural patterns of breast tissue and is an independent predictor of breast cancer risk. Although it has conceptual similarities with another breast composition measure, breast density, parenchymal texture is more complex and is represented by multiple quantitative measures. The extent to which the epidemiology and biological basis of parenchymal texture and breast density differ has not been well described. In this cross-sectional study, we evaluated quantitative measures of parenchymal texture among 308 postmenopausal women aged ≥ 45 years who participated in a routine breast cancer screening at an academic medical center between 2020 and 2022. After providing informed consent, participants completed a short interview-based questionnaire. Additional demographic and breast cancer risk factor information was obtained from the medical record. Parenchymal texture features were measured from craniocaudal views of full-field digital mammograms. Measured texture values were standardized and averaged across left and right breasts. For this analysis, we evaluated associations between features previously validated as being associated with breast cancer risk (fractal dimension, grey-level mean, co-occurrence entropy and inverse difference moment) and patient personal and clinical characteristics reported at the time of screening. Associations between texture features and participant characteristics were evaluated using Pearson correlations (continuous variables) or the Kruskal-Wallis test (categorical variables). Generalized linear models were used to adjust for technical factors associated with image acquisition, including mammography machine, compression force, and software version. P-values less than 0.05 were considered statistically significant. Among the 308 participants, 70% were White, 24% were Black, 1% were Asian, 1% were American Indian, and 4% were another race that was not specified. 4% of participants reported Hispanic ethnicity. The mean age was 64.7 years (range 45 – 84 years). Fewer than half of the women were reported by a radiologist to have dense breasts (heterogeneously dense – 24%, extremely dense – 5%, vs. almost entirely fatty – 21%, scattered fibroglandular densities – 50%). The mean body mass index (BMI) was 29.4 kg/m2 (range 19.4 to 53.1 kg/m2) and 77% of the women were parous. BMI was strongly correlated with fractal dimension (r=0.62, P< 0.01) and grey-level mean (r=-0.48, P< 0.01), and had a weaker correlation with co-occurrence entropy (r=-0.19, P< 0.01). Age was weakly correlated with co-occurrence inverse difference moment (r=-0.12, P=0.03). BI-RADS breast density was inversely associated with fractal dimension, grey-level mean, and co-occurrence entropy (all P< 0.01). Adjustment for technical factors attenuated some associations, but BMI associations with fractal dimension (P< 0.01) and gray-level mean (P< 0.01), and breast density associations with fractal dimension, grey-level mean, and co-occurrence entropy (all P< 0.01) all remained statistically significant. Parity was not associated with any of the features evaluated in unadjusted or adjusted analyses (all P >0.05). These data suggest that, in postmenopausal women, risk-associated parenchymal texture features are associated with measures affected by body size and adiposity, but not with measures related to age or reproductive status. Longitudinal studies are needed to understand the temporality of these associations. Additionally, future studies will address the joint effects of texture features and the impact of endogenous estrogen levels on the features, which will provide insight into the biological mechanisms that influence texture feature associations with personal characteristics and breast cancer risk. Citation Format: Sarah Nyante, Yukie Kajita, Walter Mankowski, Ley Killeya, Despina Kontos, Xianming Tan, Eric Cohen, Cherie Kuzmiak. Impact of patient-specific factors on quantitative breast parenchymal texture features [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO5-07-06.

MAM-E: Mammographic Synthetic Image Generation with Diffusion Models.

Generative models are used as an alternative data augmentation technique to alleviate the data scarcity problem faced in the medical imaging field. Diffusion models have gathered special attention due to their innovative generation approach, the high quality of the generated images, and their relatively less complex training process compared with Generative Adversarial Networks. Still, the implementation of such models in the medical domain remains at an early stage. In this work, we propose exploring the use of diffusion models for the generation of high-quality, full-field digital mammograms using state-of-the-art conditional diffusion pipelines. Additionally, we propose using stable diffusion models for the inpainting of synthetic mass-like lesions on healthy mammograms. We introduce MAM-E, a pipeline of generative models for high-quality mammography synthesis controlled by a text prompt and capable of generating synthetic mass-like lesions on specific regions of the breast. Finally, we provide quantitative and qualitative assessment of the generated images and easy-to-use graphical user interfaces for mammography synthesis.

Abstract 6619: Dynamic prediction with repeated mammogram images improves 5-year breast cancer risk performance

Abstract Background: Increasing use of digital screening mammography opens potential to add individualized dynamic measures of breast image history to predict future breast cancer risk. Current prediction models use a single mammogram measure, usually breast density, at one point in time, to add to prediction models based on established risk factors. Our previous work has shown that change in each breast over time is related to risk of diagnosis with breast cancer. Methods: We apply a dynamic prediction model that includes full field digital mammogram in each breast to predict future 5-year risk. Data from Joanne Knight Breast Health Cohort (JKBHC) serve as training population with 290 pathology-confirmed breast cancer cases and 658 controls sampled with a total number of 8,712 craniocaudal (CC) view mammograms for analysis. Nurses’ Health Study 2 (NHS2) serves as the external validation with 188 cases and 462 controls sampled with a total number of 3,712 CC-views. The dynamic risk prediction model aims to estimate the probability of future breast cancer risk conditional on the past history of the longitudinal mammograms. We train the model on the JKBH cohort, retaining 20% for testing and 15% for validation. We report a time dependent AUC when evaluating model performance. Results: The mean age of women was 56.6 years at entry to cohort; 81.3% were white. Adding trajectory of mammogram images in each breast to a model improved performance. The mean number of mammograms conditional on 5 years of observation in JKBHC was 3.36 (1.49) and in NHS2 2.36 (1.53). We obtained a 5-year dynamic AUC of 0.81 (0.05) in the JKBHC internal validation and 0.77 (0.06) in the NHS2 external validation. Model performance was comparable in pre- and postmenopausal women. Because breasts with high mammographic density hinder detection of tumors we stratify analyses by dense and non-dense at baseline. Among women with dense breasts at baseline (BIRADS C and D), we observe a dynamic 5-year risk AUC = 0.77. In the subgroup with less dense breasts (BIRADS A and B) we observe AUC = 0.74. Conclusions: With this approach we can accommodate the irregularly observed time-varying images and project only partial information prior to doing the prediction. Adding trajectory of mammogram history in each breast improves 5-year breast cancer risk prediction beyond models used in clinic today. Citation Format: Graham A. Colditz, Rulla M. Tamimi, Bernard Rosner, Shu Jiang. Dynamic prediction with repeated mammogram images improves 5-year breast cancer risk performance [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 6619.

Abstract B078: Artificial-intelligence-driven breast density assessment in the transition from full-field digital mammograms to digital breast tomosynthesis

Abstract Introduction: To enhance reproducibility and robustness in mammographic density assessment, various artificial intelligence (AI) models have been proposed to automatically classify mammographic images into BI-RADS density categories. Despite their promising performances, so far density AI models have been assessed primarily in traditional full-field digital mammography (FFDM) images. Our study aims to assess the potential of AI in breast density assessment in FFDM versus the newer synthetic mammography (SM) images acquired with digital breast tomosynthesis. Methods: We retrospectively analyzed negative (BI-RADS 1 or 2) routine mammographic screening exams (Selenia or Selenia Dimensions; Hologic) acquired at sites within the Barnes-Jewish/Christian (BJC) Healthcare network in St. Louis, MO from 2015 to 2018. BI-RADS breast density assessments of radiologists were obtained from BJC’s mammography reporting software (Magview 7.1). For each mammographic imaging modality, a balanced dataset of 4,000 women was selected so there were equal numbers of women in each of the four BI-RADS density categories, and each woman had at least one mediolateral oblique (MLO) and one craniocaudal (CC) view per breast in that mammographic imaging modality. Previously validated pre-processing steps were applied to all FFDM and SM images to standardize image orientation and intensity. Images were then split into training, validation, and test sets at ratios of 80%, 10%, and 10%, respectively, while maintaining the distribution of breast density categories and ensuring that all images of the same woman appear only in one set. Our AI model was based on the widely used ResNet50 architecture and was designed to accept as an input a mammographic image and predict the BI-RADS breast density category that the image belongs to. Our AI model was optimized, trained, and evaluated separately for each mammographic imaging modality. We report on the AI model’s predictive accuracy on the test set for each mammographic imaging modality, for both views as well as separately for CC and MLO; accuracy differences in FFDM versus SM were assessed via bootstrapping. Results: A batch size of 32, learning rate of e-6, and Adam optimizer were chosen as the optimal hyperparameters for our AI model. Using the same hyperparameters, the AI model demonstrated substantially higher accuracy on the test set for FFDM than for SM (FFDM: accuracy = 71% ± 4.5% versus SM: accuracy = 66% ± 4.2%; p-value<0.001 for comparison). Similar conclusion held when CC and MLO views were evaluated separately (accuracy = 72% ± 4.6% versus 66% ± 4.3% for CC; accuracy = 69% ± 4.5% versus 62% ± 4.3% for MLO; p-value<0.001 for both comparisons). Conclusions: AI performance in BI-RADS breast density assessment was significantly higher on FFDM versus SM, even under the same AI model design, dataset size and training process. Our preliminary findings suggest that further AI optimizations and adaptations may be needed as we translate AI models from FFDM to the newer SM format acquired with digital breast tomosynthesis. Citation Format: Krisha Anant, Juanita Hernandez Lopez, Debbie Bennett, Aimilia Gastounioti. Artificial-intelligence-driven breast density assessment in the transition from full-field digital mammograms to digital breast tomosynthesis [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Advances in Breast Cancer Research; 2023 Oct 19-22; San Diego, California. Philadelphia (PA): AACR; Cancer Res 2024;84(3 Suppl_1):Abstract nr B078.

Interval breast cancer: radiological surveillance in screening Egyptian population

BackgroundInterval cancers might be divided into true negative interval cancer where a new lesion is detected that no sign of disease could be detected on previous screening mammogram. For false-negative interval cancers, those missed for overt symptoms and those missed for mild or undetectable ones, this includes interpretive error as benign interval cancer (benign mimics), subtle changes, masked carcinoma or slowly growing or patient factors, such as the dense breast parenchyma. Technical failure interval cancer hampered the reader to discover the abnormality. The aim of this study was to relate the risk factors for the development of the interval breast cancer such as breast density, positive family history of breast cancer in Egyptian population. Highlight the causes of missed breast cancer in order to overcome it in the future radiological interpretation.MethodsA total of 74,546 screening mammographic examinations were performed in the radiology department at specialized breast cancer center in the period between 2015 and 2021 with about 844 pathologically proved malignant cases. Out of the 844 pathologically proven breast cancer cases, 185 breast cancer patients were interval breast cancer having reported normal examination in the previous year, 88 were true interval breast cancer and 97 were missed on mammography and detected later on. The cases were subjected to full-field digital mammogram (FFDM), complementary ultrasound, contrast-enhanced digital mammography (CEDM) and magnetic resonance imaging (MRI) in some cases, and all cases were histopathologically proven by either fine needle aspiration, core needle biopsy (CNB) or open biopsy.ResultsThe mean age of the patients was 53.5 years (range 36–83 years). The overall breast cancer detection rate was 11 per 1000 women. The 185 interval cancers were present at a rate of 3.8 per 1000 women. True negative interval breast cancers where 88 cancers were not present in retrospect on screening mammograms, 17 cases present with benign findings (benign mimics mass or calcifications) and 80 cancers were missed cancers. Analysis of the cause of missed carcinoma revealed patient-related, tumor-related, or provider-related factors. Tumor factors were the most commonly encountered, accounting for 49.5% (48 cases), then provider factors in 25.8% (25 cases) while the patient factors were the least commonly encountered in 24.7% (24 cases). Recorded positive family history found in about 35% (31cases) of the true interval breast cancers.ConclusionsAlthough mammography is the standard for detecting early breast cancer, some cancers can be missed due to various causes. Mammographic interpretation must meet high standards to reduce missed cancers. Radiologists should carefully assess screening views and order additional imaging if needed. Palpable lesions and clinical data should be further examined with ultrasound and contrast imaging if necessary. Always compare current images with previous examinations to check for any changes in lesion size. When one pathology is found, search for additional lesions.

Radiomic and deep learning characterization of breast parenchyma on full field digital mammograms and specimen radiographs: a pilot study of a potential cancer field effect.

In women with biopsy-proven breast cancer, histologically normal areas of the parenchyma have shown molecular similarity to the tumor, supporting a potential cancer field effect. The purpose of this work was to investigate relationships of human-engineered radiomic and deep learning features between regions across the breast in mammographic parenchymal patterns and specimen radiographs. This study included mammograms from 74 patients with at least 1 identified malignant tumor, of whom 32 also possessed intraoperative radiographs of mastectomy specimens. Mammograms were acquired with a Hologic system and specimen radiographs were acquired with a Fujifilm imaging system. All images were retrospectively collected under an Institutional Review Board-approved protocol. Regions of interest (ROI) of were selected from three regions: within the identified tumor, near to the tumor, and far from the tumor. Radiographic texture analysis was used to extract 45 radiomic features and transfer learning was used to extract 20 deep learning features in each region. Kendall's Tau-b and Pearson correlation tests were performed to assess relationships between features in each region. Statistically significant correlations in select subgroups of features with tumor, near to the tumor, and far from the tumor ROI regions were identified in both mammograms and specimen radiographs. Intensity-based features were found to show significant correlations with ROI regions across both modalities. Results support our hypothesis of a potential cancer field effect, accessible radiographically, across tumor and non-tumor regions, thus indicating the potential for computerized analysis of mammographic parenchymal patterns to predict breast cancer risk.

International Interobserver Variability of Breast Density Assessment

PurposeThe aim of this study was to determine variability in visually assessed mammographic breast density categorization among radiologists practicing in Indonesia, the Netherlands, South Africa, and the United States. MethodsTwo hundred consecutive 2-D full-field digital screening mammograms obtained from September to December 2017 were selected and retrospectively reviewed from four global locations, for a total of 800 mammograms. Three breast radiologists in each location (team) provided consensus density assessments of all 800 mammograms using BI-RADS® density categorization. Interreader agreement was compared using Gwet’s AC2 with quadratic weighting across all four density categories and Gwet’s AC1 for binary comparison of combined not dense versus dense categories. Variability of distribution among teams was calculated using the Stuart-Maxwell test of marginal homogeneity across all four categories and using the McNemar test for not dense versus dense categories. To compare readers from a particular country on their own 200 mammograms versus the other three teams, density distribution was calculated using conditional logistic regression. ResultsFor all 800 mammograms, interreader weighted agreement for distribution among four density categories was 0.86 (Gwet’s AC2 with quadratic weighting; 95% confidence interval, 0.85-0.88), and for not dense versus dense categories, it was 0.66 (Gwet’s AC1; 95% confidence interval, 0.63-0.70). Density distribution across four density categories was significantly different when teams were compared with one another and one team versus the other three teams combined (P < .001). Overall, all readers placed the largest number of mammograms in the scattered and heterogeneous categories. ConclusionsAlthough reader teams from four different global locations had almost perfect interreader agreement in BI-RADS density categorization, variability in density distribution across four categories remained statistically significant.

Impact of Artificial Intelligence System and Volumetric Density on Risk Prediction of Interval, Screen-Detected, and Advanced Breast Cancer.

Artificial intelligence (AI) algorithms improve breast cancer detection on mammography, but their contribution to long-term risk prediction for advanced and interval cancers is unknown. We identified 2,412 women with invasive breast cancer and 4,995 controls matched on age, race, and date of mammogram, from two US mammography cohorts, who had two-dimensional full-field digital mammograms performed 2-5.5 years before cancer diagnosis. We assessed Breast Imaging Reporting and Data System density, an AI malignancy score (1-10), and volumetric density measures. We used conditional logistic regression to estimate odds ratios (ORs), 95% CIs, adjusted for age and BMI, and C-statistics (AUC) to describe the association of AI score with invasive cancer and its contribution to models with breast density measures. Likelihood ratio tests (LRTs) and bootstrapping methods were used to compare model performance. On mammograms between 2-5.5 years prior to cancer, a one unit increase in AI score was associated with 20% greater odds of invasive breast cancer (OR, 1.20; 95% CI, 1.17 to 1.22; AUC, 0.63; 95% CI, 0.62 to 0.64) and was similarly predictive of interval (OR, 1.20; 95% CI, 1.13 to 1.27; AUC, 0.63) and advanced cancers (OR, 1.23; 95% CI, 1.16 to 1.31; AUC, 0.64) and in dense (OR, 1.18; 95% CI, 1.15 to 1.22; AUC, 0.66) breasts. AI score improved prediction of all cancer types in models with density measures (PLRT values < .001); discrimination improved for advanced cancer (ie, AUC for dense volume increased from 0.624 to 0.679, Δ AUC 0.065, P = .01) but did not reach statistical significance for interval cancer. AI imaging algorithms coupled with breast density independently contribute to long-term risk prediction of invasive breast cancers, in particular, advanced cancer.

Variations of image interpretations of radiologists from different populations in mammography and tomosynthesis with different levels of breast density.

This study aims to investigate the diagnostic performances of Australian and Shanghai-based Chinese radiologists in reading full-field digital mammogram (FFDM) and digital breast tomosynthesis (DBT) with different levels of breast density. Eighty-two Australian radiologists interpreted a 60-case FFDM set, and 29 radiologists also reported a 35-case DBT set. Sixty Shanghai radiologists read the same FFDM set, and 32 radiologists read the DBT set. The diagnostic performances of Australian and Shanghai radiologists were assessed using truth data (cancer cases were biopsy proven) and compared overall in specificity, case sensitivity, lesion sensitivity, receiver operating characteristics (ROC) area under the curve, and jack-knife free-response receiver operating characteristics (JAFROC) figure of merit, and they were stratified by case characteristics using the Mann-Whitney U test. The Spearman rank test was used to explore the association between radiologists' performances and their work experience in mammogram interpretation. There were significantly higher performances of Australian radiologists compared with Shanghai radiologists in low breast density for case sensitivity, lesion sensitivity, ROC, and JAFROC in the FFDM set ( ); in high breast density, Shanghai radiologists' performances in lesion sensitivity and JAFROC were also lower than Australian radiologists ( ). In the DBT test set, Australian radiologists performed better than Shanghai radiologists in cancer detection in both low and high breast density. The work experience of Australian radiologists was positively linked to their diagnostic performances, whereas this association was not statistically significant in Shanghai radiologists. There were significant variations in reading performances between Australian and Shanghai radiologists in FFDM and DBT across different levels of breast density, lesion types, and lesion sizes. An effective training initiative tailored to suit local readers is essential to enhancing the diagnostic accuracy of Shanghai radiologists.

The FEDRA Longitudinal Study: Repeated Volumetric Breast Density Measures and Breast Cancer Risk

Mammographic breast density (MBD) is a strong independent risk factor for breast cancer (BC). We investigated the association between volumetric MBD measures, their changes over time, and BC risk in a cohort of women participating in the FEDRA (Florence-EPIC Digital mammographic density and breast cancer Risk Assessment) study. The study was carried out among 6148 women with repeated MBD measures from full-field digital mammograms and repeated information on lifestyle habits, reproductive history, and anthropometry. The association between MBD measures (modeled as time-dependent covariates), their relative annual changes, and BC risk were evaluated by adjusted Cox models. During an average of 7.8 years of follow-up, 262 BC cases were identified. BC risk was directly associated with standard deviation increments of volumetric percent density (VPD, HR 1.37, 95%CI 1.22-1.54) and dense volume (DV, HR 1.29, 95%CI 1.18-1.41). An inverse association emerged with non-dense volume (NDV, HR 0.82, 95%CI 0.69-0.98). No significant associations emerged between annual changes in VPD, DV, NDV, and BC risk. Higher values of MBD measures, modeled as time-dependent covariates, were positively associated with increased BC risk, while an inverse association was evident for increasing NDV. No effect of annual changes in MBD emerged.

Improving the Quantitative Analysis of Breast Microcalcifications: A Multiscale Approach

Accurate characterization of microcalcifications (MCs) in 2D digital mammography is a necessary step toward reducing the diagnostic uncertainty associated with the callback of indeterminate MCs. Quantitative analysis of MCs can better identify MCs with a higher likelihood of ductal carcinoma in situ or invasive cancer. However, automated identification and segmentation of MCs remain challenging with high false positive rates. We present a two-stage multiscale approach to MC segmentation in 2D full-field digital mammograms (FFDMs) and diagnostic magnification views. Candidate objects are first delineated using blob detection and Hessian analysis. A regression convolutional network, trained to output a function with a higher response near MCs, chooses the objects which constitute actual MCs. The method was trained and validated on 435 screening and diagnostic FFDMs from two separate datasets. We then used our approach to segment MCs on magnification views of 248 cases with amorphous MCs. We modeled the extracted features using gradient tree boosting to classify each case as benign or malignant. Compared to state-of-the-art comparison methods, our approach achieved superior mean intersection over the union (0.670 ± 0.121 per image versus 0.524 ± 0.034 per image), intersection over the union per MC object (0.607 ± 0.250 versus 0.363 ± 0.278) and true positive rate of 0.744 versus 0.581 at 0.4 false positive detections per square centimeter. Features generated using our approach outperformed the comparison method (0.763 versus 0.710 AUC) in distinguishing amorphous calcifications as benign or malignant.

High-resolution synthesis of high-density breast mammograms: Application to improved fairness in deep learning based mass detection.

Computer-aided detection systems based on deep learning have shown good performance in breast cancer detection. However, high-density breasts show poorer detection performance since dense tissues can mask or even simulate masses. Therefore, the sensitivity of mammography for breast cancer detection can be reduced by more than 20% in dense breasts. Additionally, extremely dense cases reported an increased risk of cancer compared to low-density breasts. This study aims to improve the mass detection performance in high-density breasts using synthetic high-density full-field digital mammograms (FFDM) as data augmentation during breast mass detection model training. To this end, a total of five cycle-consistent GAN (CycleGAN) models using three FFDM datasets were trained for low-to-high-density image translation in high-resolution mammograms. The training images were split by breast density BI-RADS categories, being BI-RADS A almost entirely fatty and BI-RADS D extremely dense breasts. Our results showed that the proposed data augmentation technique improved the sensitivity and precision of mass detection in models trained with small datasets and improved the domain generalization of the models trained with large databases. In addition, the clinical realism of the synthetic images was evaluated in a reader study involving two expert radiologists and one surgical oncologist.

The EMory BrEast imaging Dataset (EMBED): A Racially Diverse, Granular Dataset of 3.4 Million Screening and Diagnostic Mammographic Images.

Supplemental material is available for this article. Keywords: Mammography, Breast, Machine Learning © RSNA, 2023.

Abstract A119: Comparing the variability of two validated, quantitative breast density measures used in a community-based mammography registry across race and ethnicity

Abstract Background: Breast density, an independent predictor of breast cancer risk, can be measured in multiple ways. Area-based measures quantify breast tissue in a 2-dimensional (2D) space, whereas volumetric, 3-dimensional (3D) measures incorporate information on the spatial relationship of dense and non-dense tissues that is not captured by 2D measures. Prior studies have shown that both types of measures vary by race/ethnicity. We compared the consistency of the relationship between validated 2D and 3D breast density measures across racial and ethnic groups. Methods: This study used data from 920 women aged 18+ who received a full-field digital screening mammogram at one of 6 breast imaging facilities in central North Carolina between 2017-2018. Race and ethnicity were abstracted from electronic medical records. Quantitative breast density was measured from mammograms and included (1) Area-based density measured using the Laboratory for Individualized Breast Radiodensity Assessment (LIBRA), an algorithm-based software package that produces 2D estimates of the amount of dense breast tissue and total breast area; and (2) Volumetric density measured using Volpara, which produces 3D estimates of breast fibroglandular tissue volume and total breast volume. Percent density was calculated as the ratio of dense or fibroglandular tissue to total breast area or volume. Spearman correlation coefficients (R) and 95% confidence intervals (CI) were used to quantify associations between 2D and 3D measures. Results: Participants were 33-86 years old (mean=57.9, SD=11.2) and identified as White (73%), Black (19%), Asian (3%), and Other race (6%). 5% identified as Hispanic. Percent density measures were highly correlated, with strong correlations across racial groups (White: R=0.88, 95% CI: 0.86, 0.90; Black: R=0.81, 95% CI: 0.75, 0.85; Asian: R=0.91, 95% CI: 0.81, 0.96; Other: R=0.86, 95% CI: 0.77, 0.92), and ethnic groups (Hispanic: R=0.82, 95% CI: 0.71, 0.90; non-Hispanic: R=0.88, 95% CI: 0.87, 0.90). Correlations between dense area and fibroglandular volume were lower, but were also consistent across racial groups (White: R=0.41, 95% CI: 0.35, 0.47; Black: R=0.42, 95% CI: 0.29, 0.54; Asian: R=0.33, 95% CI: -0.07, 0.64; Other: R=0.55, 95% CI: 0.32, 0.72). The correlation between dense area and fibroglandular volume was slightly stronger among Hispanic vs. non-Hispanic women (R=0.62, 95% CI: 0.41, 0.76 vs. R=0.39, 95% CI: 0.33, 0.44), although confidence intervals for the two estimates overlapped suggesting the difference was not statistically significant. Conclusions: Correlations between 2D and 3D percent density measures were consistent across racial and ethnic groups. These data suggest that use of 2D vs. 3D estimation tools is unlikely to contribute to between-study heterogeneity in racial/ethnic differences for studies of breast density in diverse populations. Generalizability for Asian and Hispanic women may be limited by the low prevalence of these groups in our sample. Studies with greater representation are needed for precise estimates in these groups. Citation Format: Brianna D. Taffe, Louise Henderson, Cherie Kuzmiak, Despina Kontos, Sarah Nyante. Comparing the variability of two validated, quantitative breast density measures used in a community-based mammography registry across race and ethnicity [abstract]. In: Proceedings of the 15th AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2022 Sep 16-19; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2022;31(1 Suppl):Abstract nr A119.