National expenditure for false-positive mammograms and breast cancer overdiagnoses estimated at $4 billion a year

Machine learning unveils an immune-related DNA methylation profile in germline DNA from breast cancer patients

BACKGROUND

There is an unmet need for precise biomarkers for early non-invasive breast cancer detection. Here, we aimed to identify blood-based DNA methylation biomarkers that are associated with breast cancer.

METHODS

DNA methylation profiling was performed for 524 Asian Chinese individuals, comprising 256 breast cancer patients and 268 age-matched healthy controls, using the Infinium MethylationEPIC array. Feature selection was applied to 649,688 CpG sites in the training set. Predictive models were built by training three machine learning models, with performance evaluated on an independent test set. Enrichment analysis to identify transcription factors binding to regions associated with the selected CpG sites and pathway analysis for genes located nearby were conducted.

RESULTS

A methylation profile comprising 51 CpGs was identified that effectively distinguishes breast cancer patients from healthy controls achieving an AUC of 0.823 on an independent test set. Notably, it outperformed all four previously reported breast cancer-associated methylation profiles. Enrichment analysis revealed enrichment of genomic loci associated with the binding of immune modulating AP-1 transcription factors, while pathway analysis of nearby genes showed an overrepresentation of immune-related pathways.

CONCLUSION

This study has identified a breast cancer-associated methylation profile that is immune-related to potential for early cancer detection.

A Semiautonomous Deep Learning System to Reduce False Positives in Screening Mammography

Purpose To evaluate the ability of a semiautonomous artificial intelligence (AI) model to identify screening mammograms not suspicious for breast cancer and reduce the number of false-positive examinations. Materials and Methods The deep learning algorithm was trained using 123 248 two-dimensional digital mammograms (6161 cancers) and a retrospective study was performed on three nonoverlapping datasets of 14 831 screening mammography examinations (1026 cancers) from two U.S. institutions and one U.K. institution (2008-2017). The stand-alone performance of humans and AI was compared. Human plus AI performance was simulated to examine reductions in the cancer detection rate, number of examinations, false-positive callbacks, and benign biopsies. Metrics were adjusted to mimic the natural distribution of a screening population, and bootstrapped CIs and P values were calculated. Results Retrospective evaluation on all datasets showed minimal changes to the cancer detection rate with use of the AI device (noninferiority margin of 0.25 cancers per 1000 examinations: U.S. dataset 1, P = .02; U.S. dataset 2, P < .001; U.K. dataset, P < .001). On U.S. dataset 1 (11 592 mammograms; 101 cancers; 3810 female patients; mean age, 57.3 years ± 10.0 [SD]), the device reduced screening examinations requiring radiologist interpretation by 41.6% (95% CI: 40.6%, 42.4%; P < .001), diagnostic examinations callbacks by 31.1% (95% CI: 28.7%, 33.4%; P < .001), and benign needle biopsies by 7.4% (95% CI: 4.1%, 12.4%; P < .001). U.S. dataset 2 (1362 mammograms; 330 cancers; 1293 female patients; mean age, 55.4 years ± 10.5) was reduced by 19.5% (95% CI: 16.9%, 22.1%; P < .001), 11.9% (95% CI: 8.6%, 15.7%; P < .001), and 6.5% (95% CI: 0.0%, 19.0%; P = .08), respectively. The U.K. dataset (1877 mammograms; 595 cancers; 1491 female patients; mean age, 63.5 years ± 7.1) was reduced by 36.8% (95% CI: 34.4%, 39.7%; P < .001), 17.1% (95% CI: 5.9%, 30.1%: P < .001), and 5.9% (95% CI: 2.9%, 11.5%; P < .001), respectively. Conclusion This work demonstrates the potential of a semiautonomous breast cancer screening system to reduce false positives, unnecessary procedures, patient anxiety, and medical expenses. Keywords: Artificial Intelligence, Semiautonomous Deep Learning, Breast Cancer, Screening Mammography Supplemental material is available for this article. Published under a CC BY 4.0 license.

Unsupervised feature correlation model to predict breast abnormal variation maps in longitudinal mammograms

Breast cancer continues to be a significant cause of mortality among women globally. Timely identification and precise diagnosis of breast abnormalities are critical for enhancing patient prognosis. In this study, we focus on improving the early detection and accurate diagnosis of breast abnormalities, which is crucial for improving patient outcomes and reducing the mortality rate of breast cancer. To address the limitations of traditional screening methods, a novel unsupervised feature correlation network was developed to predict maps indicating breast abnormal variations using longitudinal 2D mammograms. The proposed model utilizes the reconstruction process of current year and prior year mammograms to extract tissue from different areas and analyze the differences between them to identify abnormal variations that may indicate the presence of cancer. The model incorporates a feature correlation module, an attention suppression gate, and a breast abnormality detection module, all working together to improve prediction accuracy. The proposed model not only provides breast abnormal variation maps but also distinguishes between normal and cancer mammograms, making it more advanced compared to the state-of-the-art baseline models. The results of the study show that the proposed model outperforms the baseline models in terms of Accuracy, Sensitivity, Specificity, Dice score, and cancer detection rate.

Computer-extracted global radiomic features can predict the radiologists' first impression about the abnormality of a screening mammogram

OBJECTIVE

Radiologists can detect the gist of abnormal based on their rapid initial impression on a mammogram (ie, global gist signal [GGS]). This study explores (1) whether global radiomic (ie, computer-extracted) features can predict the GGS; and if so, (ii) what features are the most important drivers of the signals.

METHODS

The GGS of cases in two extreme conditions was considered: when observers detect a very strong gist (high-gist) and when the gist of abnormal was not/poorly perceived (low-gist). Gist signals/scores from 13 observers reading 4191 craniocaudal mammograms were collected. As gist is a noisy signal, the gist scores from all observers were averaged and assigned to each image. The high-gist and low-gist categories contained all images in the fourth and first quartiles, respectively. One hundred thirty handcrafted global radiomic features (GRFs) per mammogram were extracted and utilized to construct eight separate machine learning random forest classifiers (All, Normal, Cancer, Prior-1, Prior-2, Missed, Prior-Visible, and Prior-Invisible) for characterizing high-gist from low-gist images. The models were trained and validated using the 10-fold cross-validation approach. The models' performances were evaluated by the area under receiver operating characteristic curve (AUC). Important features for each model were identified through a scree test.

RESULTS

The Prior-Visible model achieved the highest AUC of 0.84 followed by the Prior-Invisible (0.83), Normal (0.82), Prior-1 (0.81), All (0.79), Prior-2 (0.77), Missed (0.75), and Cancer model (0.69). Cluster shade, standard deviation, skewness, kurtosis, and range were identified to be the most important features.

CONCLUSIONS

Our findings suggest that GRFs can accurately classify high- from low-gist images.

ADVANCES IN KNOWLEDGE

Global mammographic radiomic features can accurately predict high- from low-gist images with five features identified to be valuable in describing high-gist images. These are critical in providing better understanding of the mammographic image characteristics that drive the strength of the GGSs which could be exploited to advance breast cancer (BC) screening and risk prediction, enabling early detection and treatment of BC thereby further reducing BC-related deaths.

Burden of serious harms from diagnostic error in the USA

BACKGROUND

Diagnostic errors cause substantial preventable harms worldwide, but rigorous estimates for total burden are lacking. We previously estimated diagnostic error and serious harm rates for key dangerous diseases in major disease categories and validated plausible ranges using clinical experts.

OBJECTIVE

We sought to estimate the annual US burden of serious misdiagnosis-related harms (permanent morbidity, mortality) by combining prior results with rigorous estimates of disease incidence.

METHODS

Cross-sectional analysis of US-based nationally representative observational data. We estimated annual incident vascular events and infections from 21.5 million (M) sampled US hospital discharges (2012-2014). Annual new cancers were taken from US-based registries (2014). Years were selected for coding consistency with prior literature. Disease-specific incidences for 15 major vascular events, infections and cancers ('Big Three' categories) were multiplied by literature-based rates to derive diagnostic errors and serious harms. We calculated uncertainty estimates using Monte Carlo simulations. Validity checks included sensitivity analyses and comparison with prior published estimates.

RESULTS

Annual US incidence was 6.0 M vascular events, 6.2 M infections and 1.5 M cancers. Per 'Big Three' dangerous disease case, weighted mean error and serious harm rates were 11.1% and 4.4%, respectively. Extrapolating to all diseases (including non-'Big Three' dangerous disease categories), we estimated total serious harms annually in the USA to be 795 000 (plausible range 598 000-1 023 000). Sensitivity analyses using more conservative assumptions estimated 549 000 serious harms. Results were compatible with setting-specific serious harm estimates from inpatient, emergency department and ambulatory care. The 15 dangerous diseases accounted for 50.7% of total serious harms and the top 5 (stroke, sepsis, pneumonia, venous thromboembolism and lung cancer) accounted for 38.7%.

CONCLUSION

An estimated 795 000 Americans become permanently disabled or die annually across care settings because dangerous diseases are misdiagnosed. Just 15 diseases account for about half of all serious harms, so the problem may be more tractable than previously imagined.

A Deep Learning Decision Support Tool to Improve Risk Stratification and Reduce Unnecessary Biopsies in BI-RADS 4 Mammograms

Purpose

To evaluate the performance of a biopsy decision support algorithmic model, the intelligent-augmented breast cancer risk calculator (iBRISK), on a multicenter patient dataset.

Materials and Methods

iBRISK was previously developed by applying deep learning to clinical risk factors and mammographic descriptors from 9700 patient records at the primary institution and validated using another 1078 patients. All patients were seen from March 2006 to December 2016. In this multicenter study, iBRISK was further assessed on an independent, retrospective dataset (January 2015-June 2019) from three major health care institutions in Texas, with Breast Imaging Reporting and Data System (BI-RADS) category 4 lesions. Data were dichotomized and trichotomized to measure precision in risk stratification and probability of malignancy (POM) estimation. iBRISK score was also evaluated as a continuous predictor of malignancy, and cost savings analysis was performed.

Results

The iBRISK model's accuracy was 89.5%, area under the receiver operating characteristic curve (AUC) was 0.93 (95% CI: 0.92, 0.95), sensitivity was 100%, and specificity was 81%. A total of 4209 women (median age, 56 years [IQR, 45-65 years]) were included in the multicenter dataset. Only two of 1228 patients (0.16%) in the "low" POM group had malignant lesions, while in the "high" POM group, the malignancy rate was 85.9%. iBRISK score as a continuous predictor of malignancy yielded an AUC of 0.97 (95% CI: 0.97, 0.98). Estimated potential cost savings were more than $420 million.

Conclusion

iBRISK demonstrated high sensitivity in the malignancy prediction of BI-RADS 4 lesions. iBRISK may safely obviate biopsies in up to 50% of patients in low or moderate POM groups and reduce biopsy-associated costs.Keywords: Mammography, Breast, Oncology, Biopsy/Needle Aspiration, Radiomics, Precision Mammography, AI-augmented Biopsy Decision Support Tool, Breast Cancer Risk Calculator, BI-RADS 4 Mammography Risk Stratification, Overbiopsy Reduction, Probability of Malignancy (POM) Assessment, Biopsy-based Positive Predictive Value (PPV3) Supplemental material is available for this article. Published under a CC BY 4.0 license.See also the commentary by McDonald and Conant in this issue.

Global Radiomic Features from Mammography for Predicting Difficult-To-Interpret Normal Cases

This work aimed to investigate whether global radiomic features (GRFs) from mammograms can predict difficult-to-interpret normal cases (NCs). Assessments from 537 readers interpreting 239 normal mammograms were used to categorise cases as 120 difficult-to-interpret and 119 easy-to-interpret based on cases having the highest and lowest difficulty scores, respectively. Using lattice- and squared-based approaches, 34 handcrafted GRFs per image were extracted and normalised. Three classifiers were constructed: (i) CC and (ii) MLO using the GRFs from corresponding craniocaudal and mediolateral oblique images only, based on the random forest technique for distinguishing difficult- from easy-to-interpret NCs, and (iii) CC + MLO using the median predictive scores from both CC and MLO models. Useful GRFs for the CC and MLO models were recognised using a scree test. The CC and MLO models were trained and validated using the leave-one-out-cross-validation. The models' performances were assessed by the AUC and compared using the DeLong test. A Kruskal-Wallis test was used to examine if the 34 GRFs differed between difficult- and easy-to-interpret NCs and if difficulty level based on the traditional breast density (BD) categories differed among 115 low-BD and 124 high-BD NCs. The CC + MLO model achieved higher performance (0.71 AUC) than the individual CC and MLO model alone (0.66 each), but statistically non-significant difference was found (all p > 0.05). Six GRFs were identified to be valuable in describing difficult-to-interpret NCs. Twenty features, when compared between difficult- and easy-to-interpret NCs, differed significantly (p < 0.05). No statistically significant difference was observed in difficulty between low- and high-BD NCs (p = 0.709). GRF mammographic analysis can predict difficult-to-interpret NCs.

Evaluation of an Artificial Intelligence System for Detection of Invasive Lobular Carcinoma on Digital Mammography

Introduction Early breast cancer detection with screening mammography has been shown to reduce mortality and improve breast cancer survival. This study aims to evaluate the ability of an artificial intelligence computer-aided detection (AI CAD) system to detect biopsy-proven invasive lobular carcinoma (ILC) on digital mammography. Methods This retrospective study reviewed mammograms of patients who were diagnosed with biopsy-proved ILC between January 1, 2017, and January 1, 2022. All mammograms were analyzed using cmAssist^® (CureMetrix, San Diego, California, United States), which is an AI CAD for mammography. The AI CAD sensitivity for detecting ILC on mammography was calculated and further subdivided by lesion type, mass shape, and mass margins. To account for the within-subject correlation, generalized linear mixed models were implemented to investigate the association between age, family history, and breast density and whether the AI detected a false positive or true positive. Odds ratios, 95% confidence intervals, and p-values were also calculated. Results A total of 124 patients with 153 biopsy-proven ILC lesions were included. The AI CAD detected ILC on mammography with a sensitivity of 80%. The AI CAD had the highest sensitivity for detecting calcifications (100%), masses with irregular shape (82%), and masses with spiculated margins (86%). However, 88% of mammograms had at least one false positive mark with an average number of 3.9 false positive marks per mammogram. Conclusion The AI CAD system evaluated was successful in marking the malignancy in digital mammography. However, the numerous annotations confounded the ability to determine its overall accuracy and this reduces its potential use in real-life practice.

DeMAG predicts the effects of variants in clinically actionable genes by integrating structural and evolutionary epistatic features

Despite the increasing use of genomic sequencing in clinical practice, the interpretation of rare genetic variants remains challenging even in well-studied disease genes, resulting in many patients with Variants of Uncertain Significance (VUSs). Computational Variant Effect Predictors (VEPs) provide valuable evidence in variant assessment, but they are prone to misclassifying benign variants, contributing to false positives. Here, we develop Deciphering Mutations in Actionable Genes (DeMAG), a supervised classifier for missense variants trained using extensive diagnostic data available in 59 actionable disease genes (American College of Medical Genetics and Genomics Secondary Findings v2.0, ACMG SF v2.0). DeMAG improves performance over existing VEPs by reaching balanced specificity (82%) and sensitivity (94%) on clinical data, and includes a novel epistatic feature, the 'partners score', which leverages evolutionary and structural partnerships of residues. The 'partners score' provides a general framework for modeling epistatic interactions, integrating both clinical and functional information. We provide our tool and predictions for all missense variants in 316 clinically actionable disease genes (demag.org) to facilitate the interpretation of variants and improve clinical decision-making.

Overdiagnosis Due to Screening Mammography for Breast Cancer among Women Aged 40 Years and Over: A Systematic Review and Meta-Analysis

The current systematic review and meta-analysis was conducted to estimate the incidence of overdiagnosis due to screening mammography for breast cancer among women aged 40 years and older. A PRISMA systematic search appraisal and meta-analysis were conducted. A systematic literature search of English publications in PubMed, Web of Science, EMBASE, Scopus, and Google Scholar was conducted without regard to the region or time period. Generic, methodological, and statistical data were extracted from the eligible studies. A meta-analysis was completed by utilizing comprehensive meta-analysis software. The effect size estimates were calculated using the fail-safe N test. The funnel plot and the Begg and Mazumdar rank correlation tests were employed to find any potential bias among the included articles. The strength of the association between two variables was assessed using Kendall’s tau. Heterogeneity was measured using the I-squared (I2) test. The literature search in the five databases yielded a total of 4214 studies. Of those, 30 articles were included in the final analysis, with sample sizes ranging from 451 to 1,429,890 women. The vast majority of the articles were retrospective cohort designs (24 articles). The age of the recruited women ranged between 40 and 89 years old. The incidence of overdiagnosis due to screening mammography for breast cancer among women aged 40 years and older was 12.6%. There was high heterogeneity among the study articles (I2 = 99.993), and the pooled event rate was 0.126 (95% CI: 15 0.101–0.156). Despite the random-effects meta-analysis showing a high degree of heterogeneity among the articles, the screening tests have to allow for a certain degree of overdiagnosis (12.6%) due to screening mammography for breast cancer among women aged 40 years and older. Furthermore, efforts should be directed toward controlling and minimizing the harmful consequences associated with breast cancer screening.

Patient Cost-Sharing and Utilization of Breast Cancer Diagnostic Imaging by Patients Undergoing Subsequent Testing After a Screening Mammogram

Importance

Out-of-pocket costs (OOPCs) have been largely eliminated for screening mammography. However, patients still face OOPCs when undergoing subsequent diagnostic tests after the initial screening, which represents a potential barrier to those who require follow-up testing after initial testing.

Objective

To examine the association between the degree of patient cost-sharing and the use of diagnostic breast cancer imaging after undergoing a screening mammogram.

Design, Setting, and Participants

This retrospective cohort study used medical claims from Optum's deidentified Clinformatics Data Mart Database, a commercial claims database derived from a database of administrative health claims for members of large commercial and Medicare Advantage health plans. The large commercially insured cohort included female patients aged 40 years or older with no prior history of breast cancer undergoing a screening mammogram examination. Data were collected from January 1, 2015, to December 31, 2017, and analysis was conducted from January 2021 to September 2022.

Exposures

A k-means clustering machine learning algorithm was used to classify patient insurance plans by dominant cost-sharing mechanism. Plan types were then ranked by OOPCs.

Main Outcomes and Measures

A multivariable 2-part hurdle regression model was used to examine the association between patient OOPCs and the number and type of diagnostic breast services undergone by patients observed to undergo subsequent testing.

Results

In our sample, 230 845 women (220 023 [95.3%] aged 40 to 64 years; 16 810 [7.3%] Black, 16 398 [7.1%] Hispanic, and 164 702 [71.3%] White) underwent a screening mammogram in 2016. These patients were covered by 22 828 distinct insurance plans associated with 6 025 741 enrollees and 44 911 473 distinct medical claims. Plans dominated by coinsurance were found to have the lowest mean (SD) OOPCs ($945 [$1456]), followed by balanced plans ($1017 [$1386]), plans dominated by copays ($1020 [$1408]), and plans dominated by deductibles ($1186 [$1522]). Women underwent significantly fewer subsequent breast imaging procedures in dominantly copay (24 [95% CI, 11-37] procedures per 1000 women) and dominantly deductible (16 [95% CI, 5-28] procedures per 1000 women) plans compared with coinsurance plans. Patients from all plan types underwent fewer breast magnetic resonance imaging (MRI) scans than patients in the lowest OOPC plan (balanced, 5 [95% CI, 2-12] MRIs per 1000 women; copay, 6 [95% CI, 3-6] MRI per 100 women; deductible, 6 [95% CI, 3-9] MRIs per 1000 women.

Conclusions and Relevance

Despite policies designed to remove financial barriers to access for breast cancer screening, significant financial barriers remain for women at risk of breast cancer.

Breast Cancer Risk and Screening Mammography Frequency Among Multiethnic Women

INTRODUCTION

In 2009, the U.S. Preventive Services Task Force updated recommended mammography screening frequency from annual to biennial for average-risk women aged 50-74 years. The association between estimated breast cancer risk and mammography screening frequency was evaluated.

METHODS

A single-center retrospective cohort study was conducted among racially/ethnically diverse women, aged 50-74 years, who underwent screening mammography from 2014 to 2018. Data on age, race/ethnicity, first-degree family history of breast cancer, previous benign breast biopsies, and mammographic density were extracted from the electronic health record to calculate Breast Cancer Surveillance Consortium 5-year risk of invasive breast cancer, with a 5-year risk ≥1.67% defined as high risk. Multivariable analyses were conducted to determine the association between breast cancer risk factors and mammography screening frequency (annual versus biennial). Data were analyzed from 2020 to 2022.

RESULTS

Among 12,929 women with a mean age of 61±6.9 years, 82.7% underwent annual screening mammography, and 30.7% met high-risk criteria for breast cancer. Hispanic women were more likely to screen annually than non-Hispanic Whites (85.0% vs 79.8%, respectively), despite fewer meeting high-risk criteria. In multivariable analyses adjusting for breast cancer risk factors, high- versus low/average-risk women (OR=1.17; 95% CI=1.04, 1.32) and Hispanic versus non-Hispanic White women (OR=1.46; 95% CI=1.29, 1.65) were more likely to undergo annual mammography.

CONCLUSIONS

A majority of women continue to undergo annual screening mammography despite only a minority meeting high-risk criteria, and Hispanic women were more likely to screen annually despite lower overall breast cancer risk. Future studies should focus on the implementation of risk-stratified breast cancer screening strategies.

Fractal Dimension Analysis of Pixel Dynamic Contrast Enhanced-Magnetic Resonance Imaging Pharmacokinetic Parameters for Discrimination of Benign and Malignant Breast Lesions

PURPOSE

Breast cancer is the most frequent cancer in women worldwide. However, its diagnosis mostly depends on visual examination of radiologic images, leading to an overdiagnosis with substantial costs. Therefore, a quantitative approach such as dynamic contrast enhanced (DCE)-magnetic resonance imaging (MRI) through pharmacokinetic (PK) modeling is required for reliable analysis. As PK parameters lack information on parameter heterogeneity, texture-based analysis is required to quantify PK parameter heterogeneity. Therefore, this study focused on determining the usefulness of fractal dimension (FD) as a potential imaging biomarker of tumor heterogeneity for discriminating benign and malignant breast lesions.

METHODS

Parametric maps for PK parameters, extravasation rate of contrast agent from blood plasma to extravascular extracellular space (K^trans) and volume fraction of extravascular extracellular space (v_e), were generated for the regions of interest (ROIs) under the standard model using 18 lesions. Then, tumor ROI and pixel DCE-MRI time-course data were analyzed to extract pixel values of K^trans and v_e. For each ROI, FD values of K^trans and v_e were computed using the blanket method.

RESULTS

The FD values of K^trans for benign and malignant lesions varied from 2.96 to 3.49 and from 2.37 to 3.16, respectively, whereas FD values of v_e for benign and malignant lesions varied from 3.01 to 5.15 and 2.42 to 3.44, respectively. There were significant differences in FD values derived from K^trans parametric maps (P = .0053) and v_e parametric maps (P = .0271) between benign and malignant lesions according to the statistical analysis.

CONCLUSION

Incorporating texture heterogeneity changes in breast lesions captured by FD with quantitative DCE-MRI parameters generated under the standard model is a potential marker for prediction of malignant lesions.

Behaviour within a Clinical Trial and Implications for Mammography Guidelines

Mammography guidelines have weakened in response to evidence that mammograms diagnose breast cancers that would never eventually cause symptoms, a phenomenon called "overdiagnosis." Given concerns about overdiagnosis, instead of recommending mammograms, US guidelines encourage women aged 40-49 to get them as they see fit. To assess whether these guidelines target women effectively, I propose an approach that examines mammography behavior within an influential clinical trial that followed participants long enough to find overdiagnosis. I find that women who are more likely to receive mammograms are healthier and have higher socioeconomic status. More importantly, I find that the 20-year level of overdiagnosis is at least 3.5 times higher among women who are most likely to receive mammograms. At least 36% of their cancers are overdiagnosed. These findings imply that US guidelines encourage mammograms among healthier women who are more likely to be overdiagnosed by them. Guidelines in other countries do not.

Integrating age, BMI, and serum N-glycans detected by MALDI mass spectrometry to classify suspicious mammogram findings as benign lesions or breast cancer

While mammograms are the standard tool for breast cancer screening, there remains challenges for mammography to effectively distinguish benign lesions from breast cancers, leading to many unnecessary biopsy procedures. A blood-based biomarker could provide a minimally invasive supplemental assay to increase the specificity of breast cancer screening. Serum N-glycosylation alterations have associations with many cancers and several of the clinical characteristics of breast cancer. The current study utilized a high-throughput mass spectrometry workflow to identify serum N-glycans with differences in intensities between patients that had a benign lesion from patients with breast cancer. The overall N-glycan profiles of the two patient groups had no differences, but there were several individual N-glycans with significant differences in intensities between patients with benign lesions and ductal carcinoma in situ (DCIS). Many N-glycans had strong associations with age and/or body mass index, but there were several of these associations that differed between the patients with benign lesions and breast cancer. Accordingly, the samples were stratified by the patient's age and body mass index, and N-glycans with significant differences between these subsets were identified. For women aged 50-74 with a body mass index of 18.5-24.9, a model including the intensities of two N-glycans, 1850.666 m/z and 2163.743 m/z, age, and BMI were able to clearly distinguish the breast cancer patients from the patients with benign lesions with an AUROC of 0.899 and an optimal cutoff with 82% sensitivity and 84% specificity. This study indicates that serum N-glycan profiling is a promising approach for providing clarity for breast cancer screening, especially within the subset of healthy weight women in the age group recommended for mammograms.

Incorporating carbon into health care: adding carbon emissions to health technology assessments

At the UN Climate Change Conference 26 in Glasgow, 50 countries committed to low-carbon health services, with 14 countries further committing to net-zero carbon health services by 2050. Reaching this target will require decision makers to include carbon emissions when evaluating new and existing health technologies (tests and treatments). There is currently, however, a scarcity of data on the carbon footprint of health-care interventions, nor any means for decision makers to include and consider carbon emission health-care assessments. We therefore investigated how to integrate carbon emissions calculated by environmental life cycle assessment (LCA) into health technology assessments (HTA). HTAs are extensively used in developing clinical and policy guidelines by individual public or private payers, and by government organisations. In the first section we explain the methodological differences between environmentally extended input-output and process-based LCA. The second section outlines ways in which carbon emissions calculated by LCA could be integrated with HTAs, recognising that HTAs are done in several ways by different jurisdictions. International effort and processes will be needed to ensure that robust and comprehensive carbon footprints of commonly used health-care products are freely available. The technical and implementation challenges of incorporating carbon emissions into HTAs are considerable, but not unsurmountable. Our aim is to lay foundations for meeting these challenges.

Patient, primary care provider, and stakeholder perspectives on mammography screening frequency: lessons learned from a qualitative study

BACKGROUND

U.S. professional organizations have provided conflicting recommendations on annual vs. biennial mammography screening. Potential harms of more frequent screening include increased anxiety and costs of false positive results, including unnecessary breast biopsies and overdiagnosis.

OBJECTIVE

To characterize current practices and beliefs surrounding mammography screening frequency and perspectives on using risk-based screening to inform screening intervals.

DESIGN

Semi-structured interviews informed by the Consolidated Framework for Implementation Research (CFIR).

PARTICIPANTS

Patients, primary care providers (PCPs), third-party stakeholders (breast radiologists, radiology administrators, patient advocates).

MAIN MEASURES

Qualitative data, with a codebook developed based upon prespecified implementation science constructs.

KEY RESULTS

We interviewed 25 patients, 11 PCPs, and eight key stakeholders, including three radiologists, two radiology administrators, and three patient advocates. Most patients reported having annual mammograms, however, half believed having mammograms every two years was acceptable. Some women were worried early breast cancer would be missed if undergoing biennial screening. PCPs were equally split between recommending annual and biennial mammograms. Although PCPs were interested in using breast cancer risk models to inform screening decisions, concerns raised include time burden and lack of familiarity with breast cancer risk assessment tools. All breast radiologists believed patients should receive annual mammograms, while patient advocates and radiology administrators were split between annual vs. biennial. Radiologists were worried about missing breast cancer diagnoses when mammograms are not performed yearly. Patient advocates and radiology administrators were more open to biennial mammograms and utilizing risk-based screening.

CONCLUSIONS

Uncertainty remains across stakeholder groups regarding appropriate mammogram screening intervals. Radiologists recommend annual mammography, whereas patients and PCPs were evenly split between annual vs. biennial screening, although both favored annual screening among higher-risk women. Breast cancer risk assessment tools may help facilitate decisions about screening intervals, but face barriers to widespread implementation in the primary care setting. These results will inform future implementation strategies to adopt risk-stratified breast cancer screening.

Tumor volume doubling time estimated from digital breast tomosynthesis mammograms distinguishes invasive breast cancers from benign lesions

OBJECTIVES

The aim of this study was to determine whether lesion size metrics on consecutive screening mammograms could predict malignant invasive carcinoma versus benign lesion outcome.

METHODS

We retrospectively reviewed suspicious screen-detected lesions confirmed by biopsy to be invasive breast cancers or benign that were visible on current and in-retrospect prior screening mammograms performed with digital breast tomosynthesis from 2017 to 2020. Four experienced radiologists recorded mammogram dates, breast density, lesion type, lesion diameter, and morphology on current and prior exams. We used logistic regression models to evaluate the association of invasive breast cancer outcome with lesion size metrics such as maximum dimension, average dimension, volume, and tumor volume doubling time (TVDT).

RESULTS

Twenty-eight patients with invasive ductal carcinoma or invasive lobular carcinoma and 40 patients with benign lesions were identified. The mean TVDT was significantly shorter for invasive breast cancers compared to benign lesions (0.84 vs. 2.5 years; p = 0.0025). Patients with a TVDT of less than 1 year were shown to have an odds ratio of invasive cancer of 6.33 (95% confidence interval, 2.18-18.43). Logistic regression adjusted for age, lesion maximum dimension, and lesion volume demonstrated that shorter TVDT was the size variable significantly associated with invasive cancer outcome.

CONCLUSION

Invasive breast cancers detected on current and in-retrospect prior screening mammograms are associated with shorter TVDT compared to benign lesions. If confirmed to be sufficiently predictive of benignity in larger studies, lesions visible on mammograms which in comparison to prior exams have longer TVDTs could potentially avoid additional imaging and/or biopsy.

KEY POINTS

• We propose tumor volume doubling time as a measure to distinguish benign from invasive breast cancer lesions. • Logistic regression results summarized the utility of the odds ratio in retrospective clinical mammography data.

Quality of Informed Consent in Mammography Screening-The Polish Experience

Breast cancer is one of the leading forms of cancers in women worldwide. In Poland, it accounts for approx. 20% of all cancers diagnosed, with approximately 11,000 new cases and 5000 deaths from this disease annually. To prevent unfavourable statistics, Poland introduced free breast cancer screening programmes, available to women aged 50-69. Over a million women take advantage of this programme each year. The aim of the research was to assess the quality of consent women give prior to mammography screening and address the question of whether this quality is sufficient to make an informed choice. The study was conducted on a representative group of 600 Polish women over 50 years old (475 of them had undergone mammography screening), who agreed to take part in the study. Using the computer-assisted interview technology (CATI) method, all women were asked about their perception of breast cancer and screening and those who had undergone mammography were quizzed about the consent process. They will form the focus of this research. The validated tool contained items on both the benefits and risks of screening. The results indicate that the quality of informed consent was insufficient. A discrepancy was observed in the awareness between the benefits and risks of mammography screening. The main motivations to undergo screening were: prophylactic purposes and the free-of-charge nature of this health service. Population-based screening programmes for breast cancer should be reconsidered in terms of information policy, and the quality of informed consent should be increased.

A new perspective on breast cancer diagnostic guidelines to reduce overdiagnosis

Overdiagnosis of breast cancer, defined as diagnosing a cancer that would otherwise not cause symptoms or death in a patient's lifetime, costs U.S. health care system over $1.2 billion annually. Overdiagnosis rates, estimated to be around 10%-40%, may be reduced if indolent breast findings can be identified and followed with noninvasive imaging rather than biopsy. However, there are no validated guidelines for radiologists to decide when to choose imaging options recognizing cancer grades and types. The aim of this study is to optimize breast cancer diagnostic decisions based on cancer types using a large-scale finite-horizon Markov decision process (MDP) model with 4.6 million states to help reduce overdiagnosis. We prove the optimality of a divide-and-search algorithm that relies on tight upper bounds on the optimal decision thresholds to find an exact optimal solution. We project the high-dimensional MDP onto two lower dimensional MDPs and obtain feasible upper bounds on the optimal decision thresholds. We use real data from two private mammography databases and demonstrate our model performance through a previously validated simulation model that has been used by the policy makers to set the national screening guidelines in the United States. We find that a decision-analytical framework optimizing diagnostic decisions while accounting for breast cancer types has a strong potential to improve the quality of life and alleviate the immense costs of overdiagnosis. Our model leads to a 20 % reduction in overdiagnosis on the screening population, which translates into an annual savings of approximately $300 million for the U.S. health care system.

Prevalence and correlates of false-positive results after 3-D screening mammography among uninsured women in a community outreach program

False-positive results have been rarely investigated among uninsured minority women who undergo 3-D screening mammography. Here, we analyzed data from 21,022 women participating in the Breast Screening and Patient Navigation (BSPAN) program of North Texas with an aim to report prevalence and correlates of false-positive results after 3-D screening mammography, stratified by age. False-positives were defined as a negative diagnostic mammogram or a negative biopsy within 1 year of a positive screen. We used multivariable logistic regression to assess associations of demographic and clinical covariates and false positive results for age groups 40–49 and 50–64 years. Prevalence of false-positive results was 11.8% and 9.6% in the 40–49 and 50–64 age groups, respectively. Multivariable logistic regression demonstrated that, in the 40–49 age group, women who were non-menopausal, did not use hormone replacement therapy (HRT), and had self-reported prior mammograms had higher odds of false-positive results than those who were menopausal, used HRT and had no self-reported prior mammograms, respectively. In the 50–64 age group, women with a prior self-reported diagnostic mammogram had higher odds of false-positive results than those without a prior self-reported diagnostic mammogram. This study establishes contemporary evidence regarding prevalence and correlates of false-positive results after 3-D mammography in the unique BSPAN population, and demonstrate that use of 3-D mammography is not enough to reduce false-positive rates among uninsured women served through community outreach programs. Further research is needed to explore improved techniques to reduce false-positive rates, and ensure optimal use of scarce resources in outreach programs.

Feature fusion siamese network for breast cancer detection comparing current and prior mammograms

PURPOSE

Automatic detection of very small and non-mass abnormalities from mammogram images has remained challenging. In clinical practice for each patient, radiologists commonly not only screen the mammogram images obtained during the examination, but also compare them with previous mammogram images to make a clinical decision. To design an AI system to mimic radiologists for better cancer detection, in this work we proposed an end-to-end enhanced Siamese convolutional neural network to detect breast cancer using previous year and current year mammogram images.

METHODS

The proposed Siamese based network uses high resolution mammogram images and fuses features of pairs of previous year and current year mammogram images to predict cancer probabilities. The proposed approach is developed based on the concept of one-shot learning that learns the abnormal differences between current and prior images instead of abnormal objects, and as a result can perform better with small sample size data sets. We developed two variants of the proposed network. In the first model, to fuse the features of current and previous images, we designed an enhanced distance learning network that considers not only the overall distance, but also the pixel-wise distances between the features. In the other model, we concatenated the features of current and previous images to fuse them.

RESULTS

We compared the performance of the proposed models with those of some baseline models that use current images only (ResNet and VGG) and also use current and prior images (LSTM and vanilla Siamese) in terms of accuracy, sensitivity, precision, F1 score and AUC. Results show that the proposed models outperform the baseline models and the proposed model with the distance learning network performs the best (accuracy: 0.92, sensitivity: 0.93, precision: 0.91, specificity: 0.91, F1: 0.92 and AUC: 0.95).

CONCLUSIONS

Integrating prior mammogram images improves automatic cancer classification, specially for very small and non-mass abnormalities. For classification models that integrate current and prior mammogram images, using an enhanced and effective distance learning network can advance the performance of the models. This article is protected by copyright. All rights reserved.

Development and validation of a circulating microRNA panel for the early detection of breast cancer

BACKGROUND

Mammography is widely used for breast cancer screening but suffers from a high false-positive rate. Here, we perform the largest comprehensive, multi-center study to date involving diverse ethnic groups, for the identification of circulating miRNAs for breast cancer screening.

METHODS

This study had a discovery phase (n = 289) and two validation phases (n = 374 and n = 379). Quantitative PCR profiling of 324 miRNAs was performed on serum samples from breast cancer (all stages) and healthy subjects to identify miRNA biomarkers. Two-fold cross-validation was used for building and optimising breast cancer-associated miRNA panels. An optimal panel was validated in cohorts with Caucasian and Asian samples. Diagnostic ability was evaluated using area under the curve (AUC) analysis.

RESULTS

The study identified and validated 30 miRNAs dysregulated in breast cancer. An optimised eight-miRNA panel showed consistent performance in all cohorts and was successfully validated with AUC, accuracy, sensitivity, and specificity of 0.915, 82.3%, 72.2% and 91.5%, respectively. The prediction model detected breast cancer in both Caucasian and Asian populations with AUCs ranging from 0.880 to 0.973, including pre-malignant lesions (stage 0; AUC of 0.831) and early-stage (stages I-II) cancers (AUC of 0.916).

CONCLUSIONS

Our panel can potentially be used for breast cancer screening, in conjunction with mammography.

Artificial Intelligence (AI) for Screening Mammography, From the AI Special Series on AI Applications

Artificial intelligence (AI) applications for screening mammography are being marketed for clinical use in the interpretative domains of lesion detection and diagnosis, triage, and breast density assessment, and in the noninterpretive domains of breast cancer risk assessment, image quality control, image acquisition, and dose reduction. Evidence in support of these nascent applications, particularly for lesion detection and diagnosis, is largely based on multireader studies with cancer-enriched datasets rather than rigorous clinical evaluation aligned with the application's specific intended clinical use. This article reviews commercial AI algorithms for screening mammography that are currently available for clinical practice, their use, and evidence supporting their performance. Clinical implementation considerations, such as workflow integration, governance, and ethical issues, are also described. In addition, the future of AI for screening mammography is discussed, including the development of interpretive and noninterpretive AI applications and strategic priorities for research and development.

Reducing False-Positive Biopsies using Deep Neural Networks that Utilize both Local and Global Image Context of Screening Mammograms

Breast cancer is the most common cancer in women, and hundreds of thousands of unnecessary biopsies are done around the world at a tremendous cost. It is crucial to reduce the rate of biopsies that turn out to be benign tissue. In this study, we build deep neural networks (DNNs) to classify biopsied lesions as being either malignant or benign, with the goal of using these networks as second readers serving radiologists to further reduce the number of false-positive findings. We enhance the performance of DNNs that are trained to learn from small image patches by integrating global context provided in the form of saliency maps learned from the entire image into their reasoning, similar to how radiologists consider global context when evaluating areas of interest. Our experiments are conducted on a dataset of 229,426 screening mammography examinations from 141,473 patients. We achieve an AUC of 0.8 on a test set consisting of 464 benign and 136 malignant lesions.

Circulating microRNA breast cancer biomarker detection in patient sera with surface plasmon resonance imaging biosensor

In this article, we report for the first time, the detection of circulating miRNA as a breast cancer biomarker in patient sera using surface plasmon resonance imaging biosensor. The advantage of this approach lies in the rapid, label-free and sensitive detection. The sensor excites plasmonic resonance on the gold sensor surface and specific DNA-miRNA molecular bindings elucidate responses in the plasmonic resonance image. Experiments of detecting synthetic miRNA molecules (miR-1249) were performed and the sensor resolution was found to be 63.5 nM. The sensor was further applied to screen 17 patient serum samples from National Cancer Centre Singapore and Tan Tock Seng Hospital. Sensor intensity response was found to differ by 20% between malignant and benign cases and thus forms, a potential and an important metric in distinguishing benignity and malignancy.

Artificial intelligence for the real world of breast screening

Breast cancer screening with mammography reduces mortality in the women who attend by detecting high risk cancer early. It is far from perfect with variations in both sensitivity for the detection of cancer and very wide variations in specificity, leading to unnecessary recalls and biopsies. Over the last 12 months several papers have reported on AI algorithms that perform as well as human readers on large well curated population data sets. The nature of the test sets, the way the gold standard has been calculated, the definition of a positive call, and the statistics used all influence the results. Historically retrospective studies have not predicted the real-life performance of radiologist plus machine. So, it is important to perform prospective studies before introducing Artificial intelligence into real world breast screening.

Artificial intelligence in breast cancer screening: primary care provider preferences

BACKGROUND

Artificial intelligence (AI) is increasingly being proposed for use in medicine, including breast cancer screening (BCS). Little is known, however, about referring primary care providers' (PCPs') preferences for this technology.

METHODS

We identified the most important attributes of AI BCS for ordering PCPs using qualitative interviews: sensitivity, specificity, radiologist involvement, understandability of AI decision-making, supporting evidence, and diversity of training data. We invited US-based PCPs to participate in an internet-based experiment designed to force participants to trade off among the attributes of hypothetical AI BCS products. Responses were analyzed with random parameters logit and latent class models to assess how different attributes affect the choice to recommend AI-enhanced screening.

RESULTS

Ninety-one PCPs participated. Sensitivity was most important, and most PCPs viewed radiologist participation in mammography interpretation as important. Other important attributes were specificity, understandability of AI decision-making, and diversity of data. We identified 3 classes of respondents: "Sensitivity First" (41%) found sensitivity to be more than twice as important as other attributes; "Against AI Autonomy" (24%) wanted radiologists to confirm every image; "Uncertain Trade-Offs" (35%) viewed most attributes as having similar importance. A majority (76%) accepted the use of AI in a "triage" role that would allow it to filter out likely negatives without radiologist confirmation.

CONCLUSIONS AND RELEVANCE

Sensitivity was the most important attribute overall, but other key attributes should be addressed to produce clinically acceptable products. We also found that most PCPs accept the use of AI to make determinations about likely negative mammograms without radiologist confirmation.

A Review of Applications of Machine Learning in Mammography and Future Challenges

BACKGROUND

The aim of this study is to systematically review the literature to summarize the evidence surrounding the clinical utility of artificial intelligence (AI) in the field of mammography. Databases from PubMed, IEEE Xplore, and Scopus were searched for relevant literature. Studies evaluating AI models in the context of prediction and diagnosis of breast malignancies that also reported conventional performance metrics were deemed suitable for inclusion. From 90 unique citations, 21 studies were considered suitable for our examination. Data was not pooled due to heterogeneity in study evaluation methods.

SUMMARY

Three studies showed the applicability of AI in reducing workload. Six studies demonstrated that AI can aid in diagnosis, with up to 69% reduction in false positives and an increase in sensitivity ranging from 84 to 91%. Five studies show how AI models can independently mark and classify suspicious findings on conventional scans, with abilities comparable with radiologists. Seven studies examined AI predictive potential for breast cancer and risk score calculation. Key Messages: Despite limitations in the current evidence base and technical obstacles, this review suggests AI has marked potential for extensive use in mammography. Additional works, including large-scale prospective studies, are warranted to elucidate the clinical utility of AI.

Development of a microRNA Panel for Classification of Abnormal Mammograms for Breast Cancer

Simple Summary

Breast cancer screening by mammography suffers from high rates of false positivity, resulting in unnecessary investigative imaging and biopsies. There is an unmet need for biomarkers that can distinguish between malignant and benign breast lesions. We performed miRNA profiling on 638 patients with abnormal mammograms and 100 healthy controls. A six-miRNA panel was identified and validated in an independent cohort that had an AUC of 0.881 when differentiating between cases versus those with benign lesions or healthy individuals with normal mammograms. In addition, biomarker panel scores increased with tumor size, stage and number of lymph nodes involved. This study demonstrates that circulating miRNAs can potentially be used in conjunction with mammography to differentiate between patients with malignant and benign breast lesions.

Abstract

Mammography is extensively used for breast cancer screening but has high false-positive rates. Here, prospectively collected blood samples were used to identify circulating microRNA (miRNA) biomarkers to discriminate between malignant and benign breast lesions among women with abnormal mammograms. The Discovery cohort comprised 72 patients with breast cancer and 197 patients with benign breast lesions, while the Validation cohort had 73 and 196 cancer and benign cases, respectively. Absolute expression levels of 324 miRNAs were determined using RT-qPCR. miRNA biomarker panels were identified by: (1) determining differential expression between malignant and benign breast lesions, (2) focusing on top differentially expressed miRNAs, and (3) building panels from an unbiased search among all expressed miRNAs. Two-fold cross-validation incorporating a feature selection algorithm and logistic regression was performed. A six-miRNA biomarker panel identified by the third strategy, had an area under the curve (AUC) of 0.785 and 0.774 in the Discovery and Validation cohorts, respectively, and an AUC of 0.881 when differentiating between cases versus those with benign lesions or healthy individuals with normal mammograms. Biomarker panel scores increased with tumor size, stage and number of lymph nodes involved. Our work demonstrates that circulating miRNA signatures can potentially be used with mammography to differentiate between patients with malignant and benign breast lesions.

An Organotypic Mammary Duct Model Capturing Matrix Mechanics-Dependent Ductal Carcinoma In Situ Progression

Ductal carcinoma in situ (DCIS) is a precancerous stage breast cancer, where abnormal cells are contained within the duct, but have not invaded into the surrounding tissue. However, only 30-40% of DCIS cases are likely to progress into an invasive ductal carcinoma (IDC), while the remainder are innocuous. Since little is known about what contributes to the transition from DCIS to IDC, clinicians and patients tend to opt for treatment, leading to concerns of overdiagnosis and overtreatment. In vitro models are currently being used to probe how DCIS transitions into IDC, but many models do not take into consideration the macroscopic tissue architecture and the biomechanical properties of the microenvironment. In this study, we modeled an organotypic mammary duct as a channel molded in a collagen matrix and lined with basement membrane. By adjusting the concentration of collagen (4 and 8 mg/mL), we modulated the stiffness and morphological properties of the matrix and examined how an assortment of breast cells, including the isogenic MCF10 series that spans the range from healthy to aggressive, behaved within our model. We observed distinct characteristics of breast cancer progression such as hyperplasia and invasion. Normal mammary epithelial cells (MCF10A) formed a single-cell layer on the lumen surface, whereas the most aggressive (MCF10CA1) were several cell layers thick. The model captured collagen concentration-dependent protrusive behaviors by the MCF10A and MCF10CA1 cells, as well as a known invasive cell line (MDA-MB-231). The MCF10A and MCF10CA1 cells extended protrusions into the lower collagen concentration matrix, while the MDA-MB-231 cells fully invaded matrices of either collagen concentration but to a greater distance in the higher collagen concentration matrix. Our results show that the model can recapitulate different stages of breast cancer progression and that the MCF10 series is adaptable to physiologically relevant in vitro studies, demonstrating the potential of both the model and cell lines to elucidate key factors that may contribute to understanding the transition from DCIS to IDC. Impact statement The success of early preventative measures for breast cancer has left patients susceptible to overdiagnosis and overtreatment. Limited knowledge of factors driving an invasive transition has inspired the development of in vitro models that accurately capture this phenomenon. However, current models tend to neglect the macroscopic architecture and biomechanical properties of the mammary duct. In this study, we introduce an organotypic model that recapitulates the cylindrical geometry of the tissue and the altered stroma seen in tumor microenvironments. Our model was able to capture distinct features associated with breast cancer progression, demonstrating its potential to uncover novel insights into disease progression.

Breast cancer differential diagnosis using diffuse optical spectroscopic imaging and regression with z-score normalized data

SIGNIFICANCE

Current imaging paradigms for differential diagnosis of suspicious breast lesions suffer from high false positive rates that force patients to undergo unnecessary biopsies. Diffuse optical spectroscopic imaging (DOSI) noninvasively probes functional hemodynamic and compositional parameters in deep tissue and has been shown to be sensitive to contrast between normal and malignant tissues.

AIM

DOSI methods are under investigation as an adjunct to mammography and ultrasound that could reduce false positive rates and unnecessary biopsies, particularly in radiographically dense breasts.

METHODS

We performed a retrospective analysis of 212 subjects with suspicious breast lesions who underwent DOSI imaging. Physiological tissue parameters were z-score normalized to the patient's contralateral breast tissue and input to univariate logistic regression models to discriminate between malignant tumors and the surrounding normal tissue. The models were then used to differentiate malignant lesions from benign lesions.

RESULTS

Models incorporating several individual hemodynamic parameters were able to accurately distinguish malignant tumors from both the surrounding background tissue and benign lesions with area under the curve (AUC) ≥0.85. Z-score normalization improved the discriminatory ability and calibration of these predictive models relative to unnormalized or ratio-normalized data.

CONCLUSIONS

Findings from a large subject population study show how DOSI data normalization that accounts for normal tissue heterogeneity and quantitative statistical regression approaches can be combined to improve the ability of DOSI to diagnose malignant lesions. This improved diagnostic accuracy, combined with the modality's inherent logistical advantages of portability, low cost, and nonionizing radiation, could position DOSI as an effective adjunct modality that could be used to reduce the number of unnecessary invasive biopsies.

Screening and Selection: The Case of Mammograms

We analyze selection into screening in the context of recommendations that breast cancer screening start at age 40. Combining medical claims with a clinical oncology model, we document that compliers with the recommendation are less likely to have cancer than younger women who select into screening or women who never screen. We show this selection is quantitatively important: shifting the recommendation from age 40 to 45 results in three times as many deaths if compliers were randomly selected than under the estimated patterns of selection. The results highlight the importance of considering characteristics of compliers when making and designing recommendations.

Impact of the protein corona on nanomaterial immune response and targeting ability

Over the last decade nanomaterials have had a major impact on human health for the early detection and treatment of many diseases. The future success of clinically translatable nanomaterials lies in the combination of several functionalities to realize a personalized medical experience for patients. To maintain promises, concerns arising from toxic potential and off-target accumulation of nanomaterials must be addressed first. Upon introduction to a complex biological system (e.g., following systemic administration), nanomaterials interact with all the encountered biomolecules and form the protein corona, a complex coating of plasma proteins that provides them with a totally new biological identity. As the protein corona controls the nanomaterial behavior in vivo, a precise knowledge of the relationship between biological identity and physiological response is needed but not yet achieved. Based on impressive progress made thus far, this review critically discusses how the protein corona activates immune response and influences the targeted delivery of nanomaterials. Furthermore, we comment on emerging strategies to manipulate protein binding in order to promote formation of designer artificial coronas and achieve a desired therapeutic outcome. We conclude by debating challenges that must be overcome to obtain widespread clinical adoption of nanomaterials. This article is categorized under: Nanotechnology Approaches to Biology > Cells at the Nanoscale Toxicology and Regulatory Issues in Nanomedicine > Toxicology of Nanomaterials Therapeutic Approaches and Drug Discovery > Emerging Technologies.

Raman Spectroscopy Reveals That Biochemical Composition of Breast Microcalcifications Correlates with Histopathologic Features

Breast microcalcifications are a common mammographic finding. Microcalcifications are considered suspicious signs of breast cancer and a breast biopsy is required, however, cancer is diagnosed in only a few patients. Reducing unnecessary biopsies and rapid characterization of breast microcalcifications are unmet clinical needs. In this study, 473 microcalcifications detected on breast biopsy specimens from 56 patients were characterized entirely by Raman mapping and confirmed by X-ray scattering. Microcalcifications from malignant samples were generally more homogeneous, more crystalline, and characterized by a less substituted crystal lattice compared with benign samples. There were significant differences in Raman features corresponding to the phosphate and carbonate bands between the benign and malignant groups. In addition to the heterogeneous composition, the presence of whitlockite specifically emerged as marker of benignity in benign microcalcifications. The whole Raman signature of each microcalcification was then used to build a classification model that distinguishes microcalcifications according to their overall biochemical composition. After validation, microcalcifications found in benign and malignant samples were correctly recognized with 93.5% sensitivity and 80.6% specificity. Finally, microcalcifications identified in malignant biopsies, but located outside the lesion, reported malignant features in 65% of in situ and 98% of invasive cancer cases, respectively, suggesting that the local microenvironment influences microcalcification features. This study confirms that the composition and structural features of microcalcifications correlate with breast pathology and indicates new diagnostic potentialities based on microcalcifications assessment. SIGNIFICANCE: Raman spectroscopy could be a quick and accurate diagnostic tool to precisely characterize and distinguish benign from malignant breast microcalcifications detected on mammography.

A Circulating miRNA Signature for Stratification of Breast Lesions among Women with Abnormal Screening Mammograms

Although mammography is the gold standard for breast cancer screening, the high rates of false-positive mammograms remain a concern. Thus, there is an unmet clinical need for a non-invasive and reliable test to differentiate between malignant and benign breast lesions in order to avoid subjecting patients with abnormal mammograms to unnecessary follow-up diagnostic procedures. Serum samples from 116 malignant breast lesions and 64 benign breast lesions were comprehensively profiled for 2,083 microRNAs (miRNAs) using next-generation sequencing. Of the 180 samples profiled, three outliers were removed based on the principal component analysis (PCA), and the remaining samples were divided into training (n = 125) and test (n = 52) sets at a 70:30 ratio for further analysis. In the training set, significantly differentially expressed miRNAs (adjusted p < 0.01) were identified after correcting for multiple testing using a false discovery rate. Subsequently, a predictive classification model using an eight-miRNA signature and a Bayesian logistic regression algorithm was developed. Based on the receiver operating characteristic (ROC) curve analysis in the test set, the model could achieve an area under the curve (AUC) of 0.9542. Together, this study demonstrates the potential use of circulating miRNAs as an adjunct test to stratify breast lesions in patients with abnormal screening mammograms.

A Deep Learning Model to Triage Screening Mammograms: A Simulation Study

Background Recent deep learning (DL) approaches have shown promise in improving sensitivity but have not addressed limitations in radiologist specificity or efficiency. Purpose To develop a DL model to triage a portion of mammograms as cancer free, improving performance and workflow efficiency. Materials and Methods In this retrospective study, 223 109 consecutive screening mammograms performed in 66 661 women from January 2009 to December 2016 were collected with cancer outcomes obtained through linkage to a regional tumor registry. This cohort was split by patient into 212 272, 25 999, and 26 540 mammograms from 56 831, 7021, and 7176 patients for training, validation, and testing, respectively. A DL model was developed to triage mammograms as cancer free and evaluated on the test set. A DL-triage workflow was simulated in which radiologists skipped mammograms triaged as cancer free (interpreting them as negative for cancer) and read mammograms not triaged as cancer free by using the original interpreting radiologists' assessments. Sensitivities, specificities, and percentage of mammograms read were calculated, with and without the DL-triage-simulated workflow. Statistics were computed across 5000 bootstrap samples to assess confidence intervals (CIs). Specificities were compared by using a two-tailed t test (P < .05) and sensitivities were compared by using a one-sided t test with a noninferiority margin of 5% (P < .05). Results The test set included 7176 women (mean age, 57.8 years ± 10.9 [standard deviation]). When reading all mammograms, radiologists obtained a sensitivity and specificity of 90.6% (173 of 191; 95% CI: 86.6%, 94.7%) and 93.5% (24 625 of 26 349; 95% CI: 93.3%, 93.9%). In the DL-simulated workflow, the radiologists obtained a sensitivity and specificity of 90.1% (172 of 191; 95% CI: 86.0%, 94.3%) and 94.2% (24 814 of 26 349; 95% CI: 94.0%, 94.6%) while reading 80.7% (21 420 of 26 540) of the mammograms. The simulated workflow improved specificity (P = .002) and obtained a noninferior sensitivity with a margin of 5% (P < .001). Conclusion This deep learning model has the potential to reduce radiologist workload and significantly improve specificity without harming sensitivity. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Kontos and Conant in this issue.

A Deep Learning-Based Decision Support Tool for Precision Risk Assessment of Breast Cancer

PURPOSE

The Breast Imaging Reporting and Data System (BI-RADS) lexicon was developed to standardize mammographic reporting to assess cancer risk and facilitate the decision to biopsy. Because of substantial interobserver variability in the application of the BI-RADS lexicon, the decision to biopsy varies greatly and results in overdiagnosis and excessive biopsies. The false-positive rate from mammograms is estimated to be 7% to approximately 10% overall, but within the BI-RADS 4 category, it is greater than 70%. Therefore, we developed the Breast Cancer Risk Calculator (BRISK) to target a well-characterized and specific patient subgroup (BI-RADS 4) rather than a broad heterogeneous group in assessing breast cancer risk.

METHODS

BRISK provides a novel precise risk assessment model to reduce overdiagnosis and unnecessary biopsies. It was developed by applying natural language processing and deep learning methods on 5,147 patient records archived in the Houston Methodist systemwide data warehouse from 2006 to May 2015, including imaging and pathology reports, mammographic images, and patient demographics. Key characteristics for BI-RADS 4 patients were collected and computed to output an index measure for biopsy recommendation that is clinically relevant and informative and improves upon the traditional BI-RADS 4 scores.

RESULTS

For the validation set, we assessed data from 1,247 BI-RADS 4 patients, including mammographic images and medical reports. The BRISK model sensitivity to predict malignancy was 100%, whereas the specificity was 74%. The total accuracy of our implemented model in BRISK was 81%. Overall area under the curve was 0.93.

CONCLUSION

BRISK for abnormal mammogram uses integrative artificial intelligence technology and has demonstrated high sensitivity in the prediction of malignancy. Prospective evaluation is under way and can lead to improvement in patient-physician engagement in making informed decisions with regard to biopsy.

Impact of Artificial Intelligence on Women's Imaging: Cost-Benefit Analysis

OBJECTIVE. The purpose of this article is to identify and discuss four areas in which artificial intelligence (AI) must excel to become clinically viable: performance, time, work flow, and cost. CONCLUSION. AI holds tremendous potential for transforming the practice of radiology, but certain metrics are needed to objectively quantify its impact. As patients, physicians, hospitals, and insurance companies look for value, AI must earn a role in medical imaging.

Current Trends in Cancer Biomarker Discovery Using Urinary Metabolomics: Achievements and New Challenges

BACKGROUND

The development of effective screening methods for early cancer detection is one of the foremost challenges facing modern cancer research. Urinary metabolomics has recently emerged as a potentially transformative approach to cancer biomarker discovery owing to its noninvasive sampling characteristics and robust analytical feasibility.

OBJECTIVE

To provide an overview of new developments in urinary metabolomics, cover the most promising aspects of hyphenated techniques in untargeted and targeted metabolomics, and to discuss technical and clinical limitations in addition to the emerging challenges in the field of urinary metabolomics and its application to cancer biomarker discovery.

METHODS

A systematic review of research conducted in the past five years on the application of urinary metabolomics to cancer biomarker discovery was performed. Given the breadth of this topic, our review focused on the five most widely studied cancers employing urinary metabolomics approaches, including lung, breast, bladder, prostate, and ovarian cancers.

RESULTS

As an extension of conventional metabolomics, urinary metabolomics has benefitted from recent technological developments in nuclear magnetic resonance, mass spectrometry, gas and liquid chromatography, and capillary electrophoresis that have improved urine metabolome coverage and analytical reproducibility. Extensive metabolic profiling in urine has revealed a significant number of altered metabolic pathways and putative biomarkers, including pteridines, modified nucleosides, and acylcarnitines, that have been associated with cancer development and progression.

CONCLUSION

Urinary metabolomics presents a transformative new approach toward cancer biomarker discovery with high translational capacity to early cancer screening.

Screening Mammography Among Older Women: A Review of United States Guidelines and Potential Harms

In the United States, older women (aged ≥65 years) continue to receive routine screening mammography surveillance, despite limited evidence supporting the benefits to this subpopulation. This article reviews screening mammography guidelines and the potential harms of such screening for older women in the United States. Published guidelines and recommendations on screening mammography for older women from professional medical societies and organizations in the United States were reviewed from the mid-20th century to present. Observational data were then synthesized to present the documented harms from screening mammography among older women. In 1976, the American Cancer Society recommended to screen all women aged ≥40 years with no upper age limit. With time, other major U.S. medical societies adopted their own screening guidelines without a consensus on age of screening cessation. A population-wide screening effort has largely continued without an upper age limit and with it, a growing body of literature on the harms of screening older women. Reported harms from screening mammography procedures have included physical pain, psychological distress, excessive use of health services from overdiagnoses/false positives, and undue financial expenses. These costs are particularly pronounced among special populations with limited life expectancies such as those of very advanced age ≥80 years, long-term nursing home residents, and the cognitively impaired. When potential harms, remaining life years, and the viability of available treatments are considered, the burdens of screening mammography often outweigh the benefits for older women. For some cases, an individualized approach to recommendations would be appropriate. National guidelines should be updated to provide clear guidance for screening women of advanced age, especially those in special populations with limited life expectancies.

Breast Cyst Fluid Analysis Correlations with Speed of Sound Using Transmission Ultrasound

RATIONALE AND OBJECTIVES

The purpose of this work is to determine if the speed of sound value of a breast cyst can aid in the clinical management of breast masses. Breast macrocysts are defined as fluid-filled tissue masses >1 cm in diameter and are thought to be aberrations of normal development and involution, often associated with apocrine metaplasia. The benign natural history of breast cysts is well known, and it is important to obtain high specificity in breast imaging to avoid unnecessary biopsies in women who have benign diseases, particularly those with dense breast tissue. Transmission ultrasound is a tomographic imaging modality that generates high-resolution, 3D speed of sound maps that could be used to identify breast tissue types and act as a biomarker to differentiate lesions. We performed this study to investigate the microanatomy of macrocysts observed using transmission ultrasound, as well as assess the relationship of speed of sound to the physical and biochemical parameters of cyst fluids.

MATERIALS AND METHODS

Cyst fluid samples were obtained from 37 patients as part of a case-collection study for ultrasound imaging of the breast. The speed of sound of each sample was measured using a quantitative transmission ultrasound scanner in vivo. Electrolytes, protein, cholesterol, viscosity, and specific gravity were also measured (in the aspirated cyst fluid) to assess their relationship to the speed of sound values obtained during breast imaging.

RESULTS

We found positive correlations between viscosity and cholesterol (r = 0.71) and viscosity and total protein × cholesterol (r = 0.78). Additionally, we performed direct cell counts on cyst fluids and confirmed a positive correlation of number of cells with speed of sound (r = 0.74). The speed of sound of breast macrocysts, as observed using transmission ultrasound, correlated with the cytological features of intracystic cell clumps.

CONCLUSION

On the basis of our work with speed as a classifier, we propose a spectrum of breast macrocysts from fluid-filled to highly cellular. Our results suggest high-speed cysts are mature macrocysts with high cell counts and many cellular clumps that correlate with cyst microanatomy as seen by transmission ultrasound. Further studies are needed to confirm our findings and to assess the clinical value of speed of sound measurements in breast imaging using transmission ultrasound.

Combined Benefit of Quantitative Three-Compartment Breast Image Analysis and Mammography Radiomics in the Classification of Breast Masses in a Clinical Data Set

Purpose To investigate the combination of mammography radiomics and quantitative three-compartment breast (3CB) image analysis of dual-energy mammography to limit unnecessary benign breast biopsies. Materials and Methods For this prospective study, dual-energy craniocaudal and mediolateral oblique mammograms were obtained immediately before biopsy in 109 women (mean age, 51 years; range, 31-85 years) with Breast Imaging Reporting and Data System category 4 or 5 breast masses (35 invasive cancers, 74 benign) from 2013 through 2017. The three quantitative compartments of water, lipid, and protein thickness at each pixel were calculated from the attenuation at high and low energy by using a within-image phantom. Masses were automatically segmented and features were extracted from the low-energy mammograms and the quantitative compartment images. Tenfold cross-validations using a linear discriminant classifier with predefined feature signatures helped differentiate between malignant and benign masses by means of (a) water-lipid-protein composition images alone, (b) mammography radiomics alone, and (c) a combined image analysis of both. Positive predictive value of biopsy performed (PPV₃) at maximum sensitivity was the primary performance metric, and results were compared with those for conventional diagnostic digital mammography. Results The PPV₃ for conventional diagnostic digital mammography in our data set was 32.1% (35 of 109; 95% confidence interval [CI]: 23.9%, 41.3%), with a sensitivity of 100%. In comparison, combined mammography radiomics plus quantitative 3CB image analysis had PPV₃ of 49% (34 of 70; 95% CI: 36.5%, 58.9%; P < .001), with a sensitivity of 97% (34 of 35; 95% CI: 90.3%, 100%; P < .001) and 35.8% (39 of 109) fewer total biopsies (P < .001). Conclusion Quantitative three-compartment breast image analysis of breast masses combined with mammography radiomics has the potential to reduce unnecessary breast biopsies. © RSNA, 2018 Online supplemental material is available for this article.

Cost-effective interventions for breast cancer, cervical cancer, and colorectal cancer: new results from WHO-CHOICE

BACKGROUND

Following the adoption of the Global Action Plan for the Prevention and Control of NCDs 2013-2020, an update to the Appendix 3 of the action plan was requested by Member States in 2016, endorsed by the Seventieth World Health Assembly in May 2017 and provides a list of recommended NCD interventions. The main contribution of this paper is to present results of analyses identifying how decision makers can achieve maximum health gain using the cancer interventions listed in the Appendix 3. We also present methods used to calculate new WHO-CHOICE cost-effectiveness results for breast cancer, cervical cancer, and colorectal cancer in Southeast Asia and eastern sub-Saharan Africa.

METHODS

We used "Generalized Cost-Effectiveness Analysis" for our analysis which uses a hypothetical null reference case, where the impacts of all current interventions are removed, in order to identify the optimal package of interventions. All health system costs, regardless of payer, were included. Health outcomes are reported as the gain in healthy life years due to a specific intervention scenario and were estimated using a deterministic state-transition cohort simulation (Markov model).

RESULTS

Vaccination against human papillomavirus (two doses) for 9-13-year-old girls (in eastern sub-Saharan Africa) and HPV vaccination combined with prevention of cervical cancer by screening of women aged 30-49 years through visual inspection with acetic acid linked with timely treatment of pre-cancerous lesions (in Southeast Asia) were found to be the most cost effective interventions. For breast cancer, in both regions the treatment of breast cancer, stages I and II, with surgery ± systemic therapy, at 95% coverage, was found to be the most cost-effective intervention. For colorectal cancer, treatment of colorectal cancer, stages I and II, with surgery ± chemotherapy and radiotherapy, at 95% coverage, was found to be the most cost-effective intervention.

CONCLUSION

The results demonstrate that cancer prevention and control interventions are cost-effective and can be implemented through a step-wise approach to achieve maximum health benefits. As the global community moves toward universal health coverage, this analysis can support decision makers in identifying a core package of cancer services, ensuring treatment and palliative care for all.