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.

Recent Trends in Screening Breast MRI.

OBJECTIVE: The objective of this study was to assess trends in screening breast MRI utilization among privately insured women in the U.S. from 2007 to 2017. METHODS: The utilization of screening breast MRI among women aged 25-64 years from January 1, 2007, to December 31, 2017, was obtained using the MarketScan Commercial Database. We used Current Procedural Terminology codes to exclude breast MRI exams performed in women with a new breast cancer diagnosis and in women imaged to assess response to neoadjuvant therapy in the preceding 90 days. During the 11-year study, 351 763 study-eligible women underwent 488 852 MRI scans. RESULTS: An overall 55.0% increase in screening breast MRI utilization was observed over the study period, with a steadily increasing trend. The greatest annual increase in percent utilization was from 2007 to 2008 at 16.6%. The highest utilization rate was in 2017, in which 0.4% of women aged 25-64 years underwent screening breast MRI. Of the women who underwent screening MRI with sufficient follow-up, 76.5% underwent only one examination during the study period. CONCLUSION: Utilization of screening breast MRI has increased steadily in the past decade to a peak of 0.4% of adult women. However, an estimated 9% of U.S. women are eligible for high-risk breast MRI screening; thus, utilization falls short of optimal compliance. Further studies to evaluate the barriers to screening compliance may help optimize utilization.

A prospective cohort study to analyze the interaction of tumor-to-breast volume in breast conservation therapy versus mastectomy with reconstruction.

PURPOSE: We explored the impact of the relative volume of a tumor versus the entire breast on outcomes in patients undergoing breast conservation therapy (BCT) versus mastectomy and reconstruction (M + R). We hypothesized that there would be a threshold tumor:breast ratio (TBR) below which patient-reported outcomes (PRO) would favor BCT and above which would favor M + R. METHODS: We conducted a prospective cohort study of patients with ductal carcinoma in situ (DCIS) or invasive breast cancers undergoing BCT or M + R. A prerequisite for inclusion, analysis of tumor and breast volumes was conducted from three-dimensional magnetic resonance imaging reconstructions to calculate the TBR. Three-dimensional photography was utilized to calculate pre- and postoperative volumes and assess symmetry. Oncologic, surgical, and patient-reported outcome data were obtained from relevant BREAST-Q modules administered pre- and postoperatively. RESULTS: The BCT cohort had significantly smaller tumor volumes (p = 0.001) and lower TBRs (p = 0.001) than patients undergoing M + R overall. The M + R group, however, comprised a broader range of TBRs, characterized at lower values by patients opting for contralateral prophylactic mastectomy. Postoperative satisfaction with breasts, psychosocial, and sexual well-being scores were significantly higher in the BCT cohort, while physical well-being significantly favored the M + R cohort 480.2 ± 286.3 and 453.1 ± 392.7 days later, respectively. CONCLUSIONS: Relative to BCT, M + R was used to manage a broad range of TBRs. The relative importance of oncologic and surgical risk reduction, symmetry, and number of procedures can vary considerably and may limit the utility of TBR as a guide for deciding between BCT and M + R. Clinical Trial StatementThis study was registered with clinicaltrials.gov as "A Prospective Trial to Assess Tumor:Breast Ratio and Patient Satisfaction Following Lumpectomy Versus Mastectomy With Reconstruction", Identifier: NCT02216136.

Adiposity Change Over the Life Course and Mammographic Breast Density in Postmenopausal Women.

Mammographic breast density is a strong risk factor for breast cancer. We comprehensively investigated the associations of body mass index (BMI) change from ages 10, 18, and 30 to age at mammogram with mammographic breast density in postmenopausal women. We used multivariable linear regression models, adjusted for confounders, to investigate the associations of BMI change with volumetric percent density, dense volume, and nondense volume, assessed using Volpara in 367 women. At the time of mammogram, the mean age was 57.9 years. Compared with women who had a BMI gain of 0.1-5 kg/m2 from age 10, women who had a BMI gain of 5.1-10 kg/m2 had a 24.4% decrease [95% confidence interval (CI), 6.0%-39.2%] in volumetric percent density; women who had a BMI gain of 10.1-15 kg/m2 had a 46.1% decrease (95% CI, 33.0%-56.7%) in volumetric percent density; and women who had a BMI gain of >15 kg/m2 had a 56.5% decrease (95% CI, 46.0%-65.0%) in volumetric percent density. Similar, but slightly attenuated associations were observed for BMI gain from ages 18 and 30 to age at mammogram and volumetric percent density. BMI gain over the life course was positively associated with nondense volume, but not dense volume. We observed strong associations between BMI change over the life course and mammographic breast density. The inverse associations between early-life adiposity change and volumetric percent density suggest that childhood adiposity may confer long-term protection against postmenopausal breast cancer via its effect of mammographic breast density.

Focal breast pain: imaging evaluation and outcomes.

OBJECTIVES: To determine the number and characteristics of cancers detected and the optimal imaging evaluation in women presenting with focal breast pain (FBP). MATERIALS AND METHODS: We performed a retrospective review of 4720 women who underwent imaging for FBP from 2001 to 2013. Women 18 and over with one or two foci of breast pain and no concurrent breast symptoms were included. 944 patients met criteria. We recorded the imaging work-up, presence and type of finding at the site of pain, BI-RADS® assessment, and pathological outcomes. Subsequent imaging and clinical follow up was recorded. RESULTS: Imaging evaluation consisted of sonogram alone in 286 women, mammogram alone in 231 women, and both in 427 women. 113 women had an imaging finding at the site of pain; 103 were designated benign or probably benign. 12 biopsies of corresponding findings were performed: 9 benign, 1 invasive lobular carcinoma, 1 invasive ductal carcinoma, 1 ductal carcinoma in situ. All three malignancies were seen mammographically; 2 had an ultrasound correlate. At initial evaluation, 4 incidental breast cancers were diagnosed remote from the site of FBP. All were seen on mammogram and 2 of 4 had an ultrasound correlate. On follow up evaluation, 9 cancers were diagnosed at the site of pain and 13 incidental cancers were diagnosed. CONCLUSION: FBP is rarely associated with malignancy. Targeted ultrasound may be deferred in women 40 and older with FBP, no other clinical findings, and a negative mammogram.

Circulating Receptor Activator of Nuclear Factor-κB (RANK), RANK ligand (RANKL), and Mammographic Density in Premenopausal Women.

The receptor activator of nuclear factor-κB (RANK) pathway plays essential roles in breast development. Mammographic density is a strong risk factor for breast cancer, especially in premenopausal women. We, therefore, investigated the associations of circulating RANK and soluble RANK ligand (sRANKL) with mammographic density in premenopausal women. Mammographic density was measured as volumetric percent density in 365 cancer-free premenopausal women (mean age, 47.5 years) attending screening mammogram at the Washington University School of Medicine (St. Louis, MO). We used linear regression models adjusted for confounders, to compare the least-square means of volumetric percent density across tertiles of circulating RANK and sRANKL. Furthermore, because RANKL levels in mammary tissue are modulated by progesterone, we stratified analyses by progesterone levels. The mean volumetric percent density increased across tertiles of circulating RANK from 8.6% in tertile 1, to 8.8% in tertile 2, and 9.5% in tertile 3 (P trend = 0.02). For sRANKL, the mean volumetric percent density was 8.5% in tertile 1, 9.4% in tertile 2, and 9.0% in tertile 3 (P trend = 0.30). However, when restricted to women with higher progesterone levels, the mean volumetric percent density increased from 9.1% in sRANKL tertile 1 to 9.5% in tertile 2, and 10.1% in tertile 3 (P trend = 0.01). Circulating RANK was positively associated with volumetric percent density, while circulating sRANKL was positively associated with volumetric percent density among women with higher progesterone levels. These findings support the inhibition of RANKL signaling as a pathway to reduce mammographic density and possibly breast cancer incidence in high-risk women with dense breasts.

Single-breath-hold photoacoustic computed tomography of the breast.

We have developed a single-breath-hold photoacoustic computed tomography (SBH-PACT) system to reveal detailed angiographic structures in human breasts. SBH-PACT features a deep penetration depth (4 cm in vivo) with high spatial and temporal resolutions (255 µm in-plane resolution and a 10 Hz 2D frame rate). By scanning the entire breast within a single breath hold (~15 s), a volumetric image can be acquired and subsequently reconstructed utilizing 3D back-projection with negligible breathing-induced motion artifacts. SBH-PACT clearly reveals tumors by observing higher blood vessel densities associated with tumors at high spatial resolution, showing early promise for high sensitivity in radiographically dense breasts. In addition to blood vessel imaging, the high imaging speed enables dynamic studies, such as photoacoustic elastography, which identifies tumors by showing less compliance. We imaged breast cancer patients with breast sizes ranging from B cup to DD cup, and skin pigmentations ranging from light to dark. SBH-PACT identified all the tumors without resorting to ionizing radiation or exogenous contrast, posing no health risks.

Palpable breast abnormalities in women under age 40.

Although the prevalence of malignancy in average risk women under age 40 presenting with a palpable breast abnormality is low, the management of benign-appearing palpable abnormalities remains controversial. This study assesses the imaging evaluation, subsequent management, and outcomes of women under age 40 presenting with a palpable area of concern. This study also evaluates the costs, utility, and outcomes of BI-RADS 3 assessment in this patient population. A single institution retrospective case review from July 2010 through June 2013 identified women under age 40 presenting with a new palpable breast abnormality. Diagnostic imaging evaluation was performed. BI-RADS assessments and recommendations were recorded prospectively. Outcome was determined by tissue diagnosis, 2 years of surveillance, or search of the hospital tumor registry. Performance measures were calculated. Among 1440 cases, 1052 were initially assessed as BI-RADS 1 or 2 (73.1%), 184 as BI-RADS 3 (12.8%), 182 as BI-RADS 4 (12.6%), and 22 as BI-RADS 5 (1.5%). In all, 30 breast malignancies were diagnosed (cancer yield 2.1%). All 30 cancers were initially categorized as BI-RADS 4 or 5. No BI-RADS 1, 2, or 3 findings proved malignant. The imaging evaluation sensitivity was 100%, specificity was 87.7%, and accuracy was 87.9%. The negative predictive value was 100% and the positive predictive value was 14.7%. Average risk women under age 40 presenting with a palpable abnormality have a low prevalence of breast cancer. Imaging evaluation has a high sensitivity and negative predictive value, thereby allowing for confident characterization and appropriate management recommendations. For palpable solid masses with benign imaging features in women under age 40, short-term interval follow-up with subsequent periodic imaging or clinical examination for a total of 2 years is a cost-effective and safe alternative to biopsy.

American College of Radiology Accreditation, Performance Metrics, Reimbursement, and Economic Considerations in Breast MR Imaging.

Accreditation through the American College of Radiology (ACR) Breast Magnetic Resonance Imaging Accreditation Program is necessary to qualify for reimbursement from Medicare and many private insurers and provides facilities with peer review on image acquisition and clinical quality. Adherence to ACR quality control and technical practice parameter guidelines for breast MR imaging and performance of a medical outcomes audit program will maintain high-quality imaging and facilitate accreditation. Economic factors likely to influence the practice of breast MR imaging include cost-effectiveness, competition with lower-cost breast-imaging modalities, and price transparency, all of which may lower the cost of MR imaging and allow for greater utilization.

Adiposity at Age 10 and Mammographic Density among Premenopausal Women.

Although childhood adiposity is inversely associated with breast cancer risk, the association of childhood adiposity with mammographic density in premenopausal women has not been adequately studied. We analyzed data from 365 premenopausal women who came in for screening mammography at Washington University (St. Louis, MO) from 2015 to 2016. Body size at age 10 was self-reported using somatotype pictogram. Body mass index (BMI) at age 10 was imputed using data from Growing Up Today Study. Volpara software was used to evaluate volumetric percent density (VPD), dense volume (DV), and nondense volume (NDV). Adjusted multivariable linear regression models were used to evaluate the associations between adiposity at age 10 and mammographic density measures. Adiposity at age 10 was inversely associated with VPD and positively associated with NDV. A 1 kg/m2 increase in BMI at age 10 was associated with a 6.4% decrease in VPD, and a 6.9% increase in NDV (P < 0.001). Compared with women whose age 10 body size was 1 or 2, women with body size 3 or 4 had a 16.8% decrease in VPD and a 26.6% increase in NDV, women with body size 5 had a 32.2% decrease in VPD and a 58.5% increase in NDV, and women with body sizes ≥6 had a 47.8% decrease in VPD and a 80.9% increase in NDV (P < 0.05). The associations were attenuated, but still significant after adjusting for current BMI. Mechanistic studies to understand how childhood adiposity influences breast development, mammographic density, and breast cancer in premenopausal women are needed. Cancer Prev Res; 11(5); 287-94. ©2018 AACR.

The Future of Contrast-Enhanced Mammography.

OBJECTIVE: The purpose of this article is to discuss facilitators of and barriers to future implementation of contrast-enhanced mammography (CEM) in the United States. CONCLUSION: CEM provides low-energy 2D mammographic images analogous to digital mammography and contrast-enhanced recombined images that allow assessment of neovascularity similar to that offered by MRI. The utilization of CEM in the United States is currently low but could increase rapidly given the many potential indications for its clinical use.

Increased breast tissue receptor activator of nuclear factor-κB ligand (RANKL) gene expression is associated with higher mammographic density in premenopausal women.

Increased mammographic breast density is associated with a 4-6-fold increased risk of breast cancer, yet lifestyle factors that can reduce dense breasts are yet to be identified, and viable prevention strategies to reduce breast density-associated breast cancer development are yet to be developed. We investigated the associations of breast tissue receptor activator of nuclear factor-κB (RANK) pathway gene expression with mammographic density in 48 premenopausal women, with no previous history of cancer. Gene expression levels were measured in total RNA isolated from formalin-fixed paraffin-embedded breast tissue samples, using the NanoString nCounter platform. Mammographic density was classified based on the American College of Radiology Breast Imaging Reporting and Data (BI-RADS). Linear regression was used to evaluate associations between gene expression and mammographic density. The mean age of participants was 44.4 years. Women with higher breast tissue RANKL (TNFSF11) (p-value = 0.0076), and TNF (p-value = 0.007) gene expression had higher mammographic density. Our finding provides mechanistic support for a breast cancer chemoprevention trial with a RANKL inhibitor among high-risk premenopausal women with dense breasts.

Generation of anatomically realistic numerical phantoms for photoacoustic and ultrasonic breast imaging.

Photoacoustic computed tomography (PACT) and ultrasound computed tomography (USCT) are emerging modalities for breast imaging. As in all emerging imaging technologies, computer-simulation studies play a critically important role in developing and optimizing the designs of hardware and image reconstruction methods for PACT and USCT. Using computer-simulations, the parameters of an imaging system can be systematically and comprehensively explored in a way that is generally not possible through experimentation. When conducting such studies, numerical phantoms are employed to represent the physical properties of the patient or object to-be-imaged that influence the measured image data. It is highly desirable to utilize numerical phantoms that are realistic, especially when task-based measures of image quality are to be utilized to guide system design. However, most reported computer-simulation studies of PACT and USCT breast imaging employ simple numerical phantoms that oversimplify the complex anatomical structures in the human female breast. We develop and implement a methodology for generating anatomically realistic numerical breast phantoms from clinical contrast-enhanced magnetic resonance imaging data. The phantoms will depict vascular structures and the volumetric distribution of different tissue types in the breast. By assigning optical and acoustic parameters to different tissue structures, both optical and acoustic breast phantoms will be established for use in PACT and USCT studies.

Plasma carotenoids and the risk of premalignant breast disease in women aged 50 and younger: a nested case-control study.

PURPOSE: To examine the association of plasma carotenoids, micronutrients in fruits, and vegetables, with risk of premalignant breast disease (PBD) in younger women. METHODS: Blood samples were collected at the Siteman Cancer Center between 2008 and 2012 from 3537 women aged 50 or younger with no history of cancer or PBD. The analysis included 147 participants diagnosed with benign breast disease or breast carcinoma in situ during a 27-month follow-up and 293 controls. Cases and controls were matched on age, race/ethnicity, and date of and fasting status at blood draw. Plasma carotenoids were quantified. We used logistic regression to calculate odds ratios (ORs) and 95% confidence intervals (CIs) and linear regression to assess racial differences in plasma carotenoids. RESULTS: The risk reduction between the highest and lowest tertiles varied by carotenoid, with ß-cryptoxanthin having the greatest reduction (OR 0.62; 95% CI, 0.62-1.09; P trend = 0.056) and total carotenoids the least (OR 0.83; 95% CI, 0.48-1.44; P trend = 0.12). We observed an inverse association between plasma carotenoids and risk of PBD in obese women (BMI ≥ 30 kg/m2; 61 cases and 115 controls) but not lean women (BMI < 25 kg/m2; 54 cases and 79 controls), although the interaction was not statistically significant. Compared to white women, black women had lower levels of α and ß-carotene and higher levels of ß-cryptoxanthin and lutein/zeaxanthin. CONCLUSIONS: We observed suggestive inverse associations between plasma carotenoids and risk of PBD in younger women, consistent with inverse associations reported for invasive breast cancer. Carotenoids may play a role early in breast cancer development.

Successful Completion of the Pilot Phase of a Randomized Controlled Trial Comparing Sentinel Lymph Node Biopsy to No Further Axillary Staging in Patients with Clinical T1-T2 N0 Breast Cancer and Normal Axillary Ultrasound.

BACKGROUND: Axillary surgery is not considered therapeutic in patients with clinical T1-T2 N0 breast cancer. The importance of axillary staging is eroding in an era in which tumor biology, as defined by biomarker and gene expression profile, is increasingly important in medical decision making. We hypothesized that axillary ultrasound (AUS) is a noninvasive alternative to sentinel lymph node biopsy (SLNB), and AUS could replace SLNB without compromising patient care. STUDY DESIGN: Patients with clinical T1-T2 N0 breast cancer and normal AUS were eligible for enrollment. Subjects were randomized to no further axillary staging (arm 1) vs SLNB (arm 2). Descriptive statistics were used to describe the results of the pilot phase of the randomized controlled trial. RESULTS: Sixty-eight subjects were enrolled in the pilot phase of the trial (34 subjects in arm 1, no further staging; 32 subjects in arm 2, SLNB; and 2 subjects voluntarily withdrew from the trial). The median age was 61 years (range 40 to 80 years) in arm 1 and 59 years (range 31 to 81 years) in arm 2, and there were no significant clinical or pathologic differences between the arms. Median follow-up was 17 months (range 1 to 32 months). The negative predictive value (NPV) of AUS for identification of clinically significant axillary disease (>2.0 mm) was 96.9%. No axillary recurrences have been observed in either arm. CONCLUSIONS: Successful completion of the pilot phase of the randomized controlled trial confirms the feasibility of the study design, and provides prospective evidence supporting the ability of AUS to exclude clinically significant disease in the axilla. The results provide strong support for a phase 2 randomized controlled trial.

Predictors of false negative axillary ultrasound in breast cancer.

BACKGROUND: We sought to identify clinicopathologic factors related to false negative axillary ultrasound (AUS) results. METHODS: Patients with a clinically node-negative stage I-II breast cancer who also had a normal AUS were identified from our prospectively maintained database. All AUS studies were interpreted by dedicated breast radiologists as "normal" according to the absence of specific characteristics shown to be commonly associated with metastatic involvement. True- and false-negative AUS studies were compared statistically based on clinical, radiographic, and histologic parameters. RESULTS: Of the 118 patients with a normal AUS, 25 (21%) were ultimately found to be node-positive on pathologic assessment after axillary surgery. On bivariate analysis, primary tumor size and lymphovascular invasion (LVI) were found to be significantly different between true- and false-negative AUS. The average tumor size was smaller in the true-negative group compared with that in the false-negative group (16 versus 21 mm [P < 0.01]). The presence of LVI was more likely in the false-negative group (44%) compared with that in the true-negative group (8%, P < 0.0001). No significant difference was noted between groups with regard to patient age, race, body mass index, tumor grade, histologic type, hormone receptor status, and time between AUS and axillary surgery. On multivariate analysis, only the presence of LVI achieved statistical significance (P = 0.0007). CONCLUSIONS: AUS is a valuable tool that accurately predicted absence of axillary disease in 79% of patients with clinically node-negative breast cancer. AUS findings may be less accurate in the setting of LVI, and a negative AUS in patients with LVI should be interpreted cautiously.

Recent trends in adherence to continuous screening for breast cancer among Medicare beneficiaries.

OBJECTIVE: The aim of this study is to examine recent trends in adherence to continuous screening, especially the rate of subsequent screening mammography following an initial screening before and after the U.S. Preventive Services Task Force (USPSTF) revised its guidelines on breast cancer in November 2009. METHODS: We retrospectively analyzed Medicare fee-for-service claims data to: 1) compare rate of subsequent screening mammography over 27 month periods for 317,150 women screened in either 2004 or 2009; and 2) examine patterns of subsequent screening by age and race. RESULTS: When adjusted for age, race, state of residence, county-level covariates, and clustered on ordering provider, the rate of subsequent screening decreased in 2009 relative to 2004 (OR=0.75; 95% CI: 0.74-0.76). Adjusted odds ratios are similar for alternative follow-up windows (15 months, 0.71; 24 months, 0.70; 30 months 0.75). The decline was mostly attributable to women 75 and older who are now less likely to return for a subsequent screening. Although USPSTF guidelines call for 24 months, approximately half of women continue screening at 12-month intervals in both cohorts. CONCLUSIONS: The rate of subsequent screening mammography has declined after 2009. Older women seem to follow the revised USPSTF guideline, but confusion by physicians and patients about competing guidelines may be contributing to these findings.