Strategies for Decreasing Screening Mammography Recall Rates While Maintaining Performance Metrics

Breast Cancer Screening Using Inverse Modeling of Surface Temperatures and Steady-State Thermal Imaging

Cancer is characterized by increased metabolic activity and vascularity, leading to temperature changes in cancerous tissues compared to normal cells. This study focused on patients with abnormal mammogram findings or a clinical suspicion of breast cancer, exclusively those confirmed by biopsy. Utilizing an ultra-high sensitivity thermal camera and prone patient positioning, we measured surface temperatures integrated with an inverse modeling technique based on heat transfer principles to predict malignant breast lesions. Involving 25 breast tumors, our technique accurately predicted all tumors, with maximum errors below 5 mm in size and less than 1 cm in tumor location. Predictive efficacy was unaffected by tumor size, location, or breast density, with no aberrant predictions in the contralateral normal breast. Infrared temperature profiles and inverse modeling using both techniques successfully predicted breast cancer, highlighting its potential in breast cancer screening.

Mammography Information Systems: A Survey of Breast Imaging Radiologist Satisfaction and Perspectives

OBJECTIVE

To assess use of mammography information systems (MISs) and explore features associated with breast imaging radiologist satisfaction.

METHODS

A 22-question survey regarding MISs was distributed electronically to the Society of Breast Imaging membership between February 16, 2022 and June 28, 2022. Differences in responses between respondents satisfied and dissatisfied with their MIS were analyzed using Pearson chi-squared test, Fisher exact test, and multivariate logistic regression.

RESULTS

The response rate was 11.4% (228/2007). Most respondents used a commercial MIS (195/228, 85.5%). Most used were Epic (47/228, 21%), MagView (47/228, 21%), and PenRad (37/228, 16%). Only 4.4% (10/228) reported that patient tracking was not integrated with results reporting. The majority (129/226, 57%) reported satisfaction with their MIS. Satisfaction correlated (P < 0.05) with features such as picture archiving and communication system integration, structured reporting, access to physician outcomes metrics, and ability to query data. Less commonly reported features such as non-English language options and recognition of laterality and patient mismatch errors also correlated with satisfaction. Lack of these features correlated with dissatisfaction (P < 0.05). Satisfaction also correlated with adequate training (P < 0.001) and technology support (P < 0.001). On multivariate analysis, longer time using the current MIS was independently associated with satisfaction.

CONCLUSION

Most respondents used a commercial MIS and were satisfied with their system. Satisfied users reported several helpful MIS features and adequate training and support. The survey results could help MIS companies when designing new products and inform radiologists and administrators when considering a new MIS.

Multicenter, Multivendor Validation of an FDA-approved Algorithm for Mammography Triage

OBJECTIVE

Artificial intelligence (AI)-based triage algorithms may improve cancer detection and expedite radiologist workflow. To this end, the performance of a commercial AI-based triage algorithm on screening mammograms was evaluated across breast densities and lesion types.

METHODS

This retrospective, IRB-exempt, multicenter, multivendor study examined 1255 screening 4-view mammograms (400 positive and 855 negative studies). Images were anonymized by providing institutions and analyzed by a commercially available AI algorithm (cmTriage, CureMetrix, La Jolla, CA) that performed retrospective triage at the study level by flagging exams as "suspicious" or not. Sensitivities and specificities with confidence intervals were derived from area under the curve (AUC) calculations.

RESULTS

The algorithm demonstrated an AUC of 0.95 (95% CI: 0.94-0.96) for case identification. Area under the curve held across densities (0.95) and lesion types (masses: 0.94 [95% CI: 0.92-0.96] or microcalcifications: 0.97 [95% CI: 0.96-0.99]). The algorithm has a default sensitivity of 93% (95% CI: 95.6%-90.5%) with specificity of 76.3% (95% CI: 79.2%-73.4%). To evaluate real-world performance, a sensitivity of 86.9% (95% CI: 83.6%-90.2%) was tested, as observed for practicing radiologists by the Breast Cancer Surveillance Consortium (BCSC) study. The resulting specificity was 88.5% (95% CI: 86.4%-90.7%), similar to the BCSC specificity of 88.9%, indicating performance comparable to real-world results.

CONCLUSION

When tested for lesion detection, an AI-based triage software can perform at the level of practicing radiologists. Drawing attention to suspicious exams may improve reader specificity and help streamline radiologist workflow, enabling faster turnaround times and improving care.

Patterns of Screening Recall Behavior Among Subspecialty Breast Radiologists

RATIONALE AND OBJECTIVES

Determine whether there are patterns of lesion recall among breast imaging subspecialists interpreting screening mammography, and if so, whether recall patterns correlate to morphologies of screen-detected cancers.

MATERIALS AND METHODS

This Institutional Review Board-approved, retrospective review included all screening examinations January 3, 2012-October 1, 2018 interpreted by fifteen breast imaging subspecialists at a large academic medical center and two outpatient imaging centers. Natural language processing identified radiologist recalls by lesion type (mass, calcifications, asymmetry, architectural distortion); proportions of callbacks by lesion types were calculated per radiologist. Hierarchical cluster analysis grouped radiologists based on recall patterns. Groups were compared to overall practice and each other by proportions of lesion types recalled, and overall and lesion-specific positive predictive value-1 (PPV1).

RESULTS

Among 161,859 screening mammograms with 13,086 (8.1%) recalls, Hierarchical cluster analysis grouped 15 radiologists into five groups. There was substantial variation in proportions of lesions recalled: calcifications 13%-18% (Chi-square 45.69, p < 0.00001); mass 16%-44% (Chi-square 498.42, p < 0.00001); asymmetry 13%-47% (Chi-square 660.93, p < 0.00001) architectural distortion 6%-20% (Chi-square 283.81, p < 0.00001). Radiologist groups differed significantly in overall PPV1 (range 5.6%-8.8%; Chi-square 17.065, p = 0.0019). PPV1 by lesion type varied among groups: calcifications 9.2%-15.4% (Chi-square 2.56, p = 0.6339); mass 5.6%-8.5% (Chi-square 1.31, p = 0.8597); asymmetry 3.4%-5.9% (Chi-square 2.225, p = 0.6945); architectural distortion 5.6%-10.8% (Chi-square 5.810, p = 0.2138). Proportions of recalled lesions did not consistently correlate to proportions of screen-detected cancer.

CONCLUSION

Breast imaging subspecialists have patterns for screening mammography recalls, suggesting differential weighting of imaging findings for perceived malignant potential. Radiologist recall patterns are not always predictive of screen-detected cancers nor lesion-specific PPV1s.

Transitioning to Practice: Getting up to Speed in Efficiency and Accuracy

The transition from trainee to breast radiologist is challenging. The many new responsibilities that breast radiologists acquire while establishing themselves as clinicians may increase stress and anxiety. Taking inventory of existing knowledge and skills and addressing deficits toward the end of one's training can be beneficial. New breast radiologists should expect to be slower and gain proficiency in the first several years out of training. Having realistic expectations for oneself with respect to screening mammography interpretation and following up on the subsequent diagnostic imaging workup of screening callback examinations can increase competence and confidence. Familiarity with the available literature to guide management in the diagnostic setting can increase efficiency. Planning ahead for localizations and biopsies also allows for efficiency while alleviating anxiety. Ultimately, adapting to a new work environment using a collaborative approach with primary healthcare providers, pathologists, and surgeons while remembering to have mentors within and beyond the field of radiology allows for a more successful transition.

Incorporating Peer Learning Into Your Breast Imaging Practice

Traditional score-based peer review has come under scrutiny in recent years, as studies have demonstrated it to be generally ineffective at improving quality. Many practices and programs are transitioning to a peer learning model to replace or supplement traditional peer review. Peer learning differs from traditional score-based peer review in that the emphasis is on sharing learning opportunities and creating an environment that fosters discussion of errors in a nonpunitive forum with the goal of improved patient care. Creating a just culture is central to fostering successful peer learning. In a just culture, mistakes can be discussed without shame or fear of retribution and the focus is on systems improvement rather than individual blame. Peer learning, as it pertains to breast imaging, can occur in many forms and venues. Examples of the various formats in which peer learning can occur include through individual colleague interaction, as well as divisional, multidisciplinary, department-wide, and virtual conferences, and with the assistance of artificial intelligence. Incorporating peer learning into the practice of breast imaging aims to reduce delayed diagnoses of breast cancer and optimize patient care.

Benefits of digital breast tomosynthesis: A lesion-level analysis

OBJECTIVE

To compare outcome metrics of digital breast tomosynthesis (DBT) breast cancer screening with full-field digital mammogram (FFDM); specifically, to compare recall rates by the type of recalled finding, and to assess if screening with DBT versus FFDM changes biopsy recommendations and if the likelihood of malignancy varied by lesion type, if detected on DBT or FFDM screening mammogram.

METHODS

The outcomes of 22,055 FFDM and DBT screening mammograms were retrospectively reviewed. The exams were performed at an academic institution between August 2015 and September 2016. Performance of screening with FFDM versus DBT was compared in terms of recall rate and percentage of recalled lesions resulting in a cancer diagnosis, with subset analyses performed for specific mammographic findings.

RESULTS

The recall rate was 10.6% for FFDM and 8.0% for DBT (p < 0.001). Architectural distortion was more likely to be recalled on DBT screening than FFDM (p = 0.002), and was associated with an increased likelihood of malignancy (p = 0.008). Asymmetries were less likely to be recalled on DBT than FFDM (p < 0.001) screening mammogram, but more likely to be recommended for biopsy when detected on DBT. Calcifications more frequently required short-term follow-up or biopsy on both DBT and FFDM.

CONCLUSIONS

DBT screening confers an advantage in detection of architectural distortion representing malignancy. Recall rate of asymmetries are reduced with screening DBT, probably due to reduction of tissue superimposition. Calcifications pose a particularly difficult diagnostic challenge for breast imagers, regardless of screening mammogram type.

The Adoption and Impact on Performance of an Automated Outcomes Feedback Application for Tomosynthesis Screening Mammography

OBJECTIVE

To evaluate a tomosynthesis screening mammography automated outcomes feedback application's adoption and impact on performance.

METHODS

This prospective intervention study evaluated a feedback application that provided mammographers subsequent imaging and pathology results for patients that radiologists had personally recalled from screening. Deployed to 13 academic and 5 private practice attending radiologists, adoption was studied from March 29, 2018, to March 20, 2019. Radiologists indicated if reviewed feedback would influence future clinical decisions. For a subset of eight academic radiologists consistently interpreting screening mammograms during the study, performance metrics were compared pre-intervention (January 1, 2016, to September 30, 2017) and post-intervention (October 1, 2017 to June 30, 2018). Abnormal interpretation rate, positive predictive value of biopsies performed, sensitivity, specificity, and cancer detection rate were compared using Pearson's χ² test. Logistic regression models were fit, adjusting for age, race, breast density, prior comparison, breast cancer history, and radiologist.

RESULTS

The 18 radiologists reviewed 68.5% (1,398 of 2,042) of available feedback cases and indicated that 17.4% of cases (243 of 1,398) could influence future decisions. For the eight academic radiologist subset, after multivariable adjustment with comparison to pre-intervention, average abnormal interpretation rate decreased (from 7.5% to 6.7%, adjusted odds ratio [aOR] 0.86, P < .01), positive predictive value of biopsies performed increased (from 40.6% to 51.3%, aOR 1.48, P = .011), and specificity increased (from 93.0% to 93.9%, aOR 1.17, P < .01) post-intervention. There was no difference in cancer detection rate per 1,000 examinations (from 5.8 to 6.1, aOR 1.01, P = .91) or sensitivity (from 81.2% to 78.7%, aOR 0.84, P = .47).

CONCLUSIONS

Radiologists used a screening mammography automated outcomes feedback application. Its use decreased false-positive examinations, without evidence of reduced cancer detection.

Breast Cancer Screening with Digital Breast Tomosynthesis: Are Initial Benefits Sustained?

Background Performance metrics with digital breast tomosynthesis (DBT) are based on early experiences. There is limited research on whether the benefits of DBT are sustained. Purpose To determine whether improved screening performance metrics with DBT are sustained over time at the population level and after the first screening round at the individual level. Materials and Methods A retrospective review was conducted of screening mammograms that had been obtained before DBT implementation (March 2008 to February 2011, two-dimensional digital mammography [DM] group) and for 5 years after implementation (January 2013 to December 2017, DBT1-DBT5 groups, respectively). Patients who underwent DBT were also categorized according to the number of previous DBT examinations they had undergone. Performance metrics were compared between DM and DBT groups and between patients with no previous DBT examinations and those with at least one prior DBT examination by using multivariable logistic regression models. Results The DM group consisted of 99 582 DM examinations in 55 086 women (mean age, 57.3 years ± 11.6 [standard deviation]). The DBT group consisted of 205 048 examinations in 76 276 women (mean age, 58.2 years ± 11.2). There were no differences in the cancer detection rate (CDR) between DM and DBT groups (4.6-5.8 per 1000 examinations, P = .08 to P = .95). The highest CDR was observed with a woman's first DBT examination (6.1 per 1000 examinations vs 4.4-5.7 per 1000 examinations with at least one prior DBT examination, P = .001 to P = .054). Compared with the DM group, the DBT1 group had a lower abnormal interpretation rate (AIR) (adjusted odds ratio [AOR], 0.85; P < .001), which remained reduced in the DBT2, DBT3, and DBT5 groups (P < .001 to P = .02). The reduction in AIR was also sustained after the first examination (P < .001 to P = .002). Compared with the DM group, the DBT1 group had a higher specificity (AOR, 1.20; P < .001), which remained increased in DBT2, DBT3, and DBT5 groups (P < .001 to P = .004). The increase in specificity was also sustained after the first examination (P < .001 to P = .01). Conclusion The benefits of reduced false-positive examinations and higher specificity with screening tomosynthesis were sustained after the first screening round at the individual level. © RSNA, 2020 See also the editorial by Taourel in this issue.

Factors Impacting False Positive Recall in Screening Mammography

RATIONALE AND OBJECTIVES

Our objective was to identify factors impacting false positive recalls in screening mammography.

MATERIALS AND METHODS

We retrospectively reviewed our screening mammography database from August 31, 2015 to September 30, 2016, including full field digital mammograms (FFDM) and digital breast tomosynthesis (DBT) mammograms. False positive (FP) exams were defined as Breast Imaging-Reporting and Data System (BI-RADS) 1 or 2 assessments at diagnostic imaging with 1 year cancer-free follow-up, Breast Imaging-Reporting and Data System 3 assessment at diagnostic imaging with 2 years cancer free follow-up, or biopsy with benign pathology. True positives were defined as malignant pathology on biopsy or surgical excision. We evaluated the association of FP recalls with multiple patient-level factors and imaging features.

RESULTS

A total of 22,055 screening mammograms were performed, and 1887 patients were recalled (recall rate 8.6%). Recall rate was lower for DBT than full field digital mammograms (8.0% vs 10.6%, p < 0.001). FP results were lower if prior mammograms were available (90.8% vs 95.8%, p = 0.02), and if there was a previous benign breast biopsy (87.6% vs 92.9%, p = 0.01). Mean age for the FP group was lower than the true positive group (56.1 vs 62.9 years, p < 0.001). There were no significant differences in FP recalls based on history of high-risk lesions, family history of breast or ovarian cancer, hormone use, breast density, race, or body mass index.

CONCLUSION

FP recalls were significantly less likely with DBT, in older women, in patients with prior mammograms available for comparison, and in patients with histories of benign breast biopsy. This study supports the importance of using DBT in the screening setting and obtaining prior mammograms for comparison.

Utility of Supplemental Training to Improve Radiologist Performance in Breast Cancer Screening: A Literature Review

PURPOSE

The authors evaluate whether supplemental training for radiologists improves their breast screening performance and how this is measured.

METHODS

A systematic search was conducted in PubMed on August 3, 2017. Articles were included if they described supplemental training for radiologists reading mammograms to improve their breast screening performance and at least one outcome measure was reported. Study quality was assessed using the Medical Education Research Study Quality Instrument.

RESULTS

Of 2,199 identified articles, 18 were included, of which 17 showed improvement on at least one of the outcome measures, for at least one training activity or subgroup. Two measurement approaches were found. For the first approach, measuring performance on test sets, sensitivity, and specificity were the most reported outcomes (8 of 11 studies). Recall rate is the most reported outcome (6 of 7 studies) for the second approach, which measures performance in actual screening practice. The studies were mainly of moderate quality (Medical Education Research Study Quality Instrument score 11.7 ± 1.7), caused by small sample sizes and the lack of a control group.

CONCLUSIONS

Supplemental training helps radiologists improve their screening performance, despite the mainly moderate quality of the studies. There is a need for better designed studies. Future studies should focus on performance in actual screening practice and should look for methods to isolate the training effect. If test sets are used, focus should be on knowledge about correlation between performance on test sets and actual screening practice.

Variability in Individual Radiologist BI-RADS 3 Usage at a Large Academic Center: What's the Cause and What Should We Do About It?

RATIONALE AND OBJECTIVES

Although the breast imaging reporting and data system (BI-RADS) lists specific criteria for designating a lesion as BI-RADS category 3 (probably benign), there are no target benchmarks for BI-RADS 3 usage rates. This study investigates the variability of BI-RADS 3 rates among a group of academic breast imagers, with the goal of defining more precise utilization.

MATERIALS AND METHODS

We retrospectively reviewed all diagnostic mammograms performed between July 1, 2013 and August 8, 2017 at our academic institution. The percentage of diagnostic mammograms given a BI-RADS 3 assessment was compared between radiologists using the Chi-square test. We then evaluated for correlation between BI-RADS 3 rate and individual clinical metrics (eg, radiologist experience, cancer detection rate [CDR] and recall rate) using univariate linear regression.

RESULTS

The study included 13 breast imagers and 24,051 diagnostic breast examinations. There was significant variability in BI-RADS 3 rates between radiologists, ranging from 8.0% to 19.3% (p < 0.001). Increased BI-RADS 3 rates negatively correlated with BI-RADS 1 or 2 rate (p < 0.001) and positively correlated with recall rate (p = 0.03). There was no association between BI-RADS 3 rate and the radiologist's level of experience, BI-RADS 4 or 5 rate, or CDR.

CONCLUSION

We found significant variability in BI-RADS 3 usage, which seems to be used in place of BI-RADS 1 or 2 findings rather than to avoid biopsy recommendation. BI-RADS 3 rates also directly correlated with recall rate, suggesting a greater degree of uncertainty among specific radiologists. Importantly, increased usage of BI-RADS 3 did not correlate with provider experience or improved CDR.

The Potential Impact of Digital Breast Tomosynthesis on the Benign Biopsy Rate in Women Recalled within the UK Breast Screening Programme

Background Multiple studies showed digital breast tomosynthesis (DBT) has a higher cancer detection rate and a lower recall rate than full-field digital mammography (FFDM). However, there is a paucity of studies on whether the benign biopsy rate will be lower with DBT. Purpose To evaluate the benign biopsy rate of DBT versus that of FFDM in women recalled after breast screening within the National Health Service Breast Screening Programme. Materials and Methods This prospective single institution study included women who underwent screening FFDM or screening breast MRI between November 13, 2015, and July 29, 2016, and were recalled. Women who were recalled underwent a breast examination, additional imaging (mammography, US), and biopsy, if required. The number of interventions and the outcome of assessment were recorded. FFDM was performed at the screening appointment, and DBT images were acquired when participants attended a screening assessment within 3 weeks after screening FFDM. Two of four readers blinded to biopsy results analyzed DBT images in isolation and then in conjunction with FFDM images and any additional two-dimensional mammographic or US images. Two-tailed McNemar tests were used to test differences in sensitivity and specificity of the two conditions at 5% significance level. Results During the study period, 30 933 women underwent FFDM. A total of 1470 women were recalled (recall rate, 4.8%), and 827 were included after exclusion criteria were applied. Their mean age was 56.7 years ± 7.7 (standard deviation). A total of 145 breast cancers were detected (142 with FFDM, two with DBT only, one with surveillance MRI). Triple assessment without DBT resulted in 571 breast biopsies and enabled detection of 142 cancers. The addition of DBT would have resulted in 298 biopsies and detection of 142 cancers, reducing the number of biopsies from 571 of 827 (69.0%) to 298 of 827 (36.0%). Conclusion When compared with full-field digital mammography, digital breast tomosynthesis can reduce the benign biopsy rate while maintaining the cancer detection rate. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Skaane in this issue.

Quantifying performance thresholds for recommending screening mammography: a revealed preference analysis of USPSTF guidelines

PURPOSE

During ongoing controversies about mammography screening, many investigators have stated that performance improvements in screening mammography may mitigate concerns about harms. However, there have been few attempts to quantify performance improvements required to recommend mammography screening. Based on USPSTF benchmarks, we utilized revealed preference methods to ascertain quantitative thresholds at which screening mammography would be recommended beyond biennial screening in women 50 and older.

METHODS

Benefits of routine screening mammography (breast cancer deaths averted) were from published USPSTF meta-analyses. Potential harms (10-year cumulative probability of at least one false-positive) were from published Breast Cancer Surveillance Consortium estimates. We identified the implicit threshold (benefit/harm ratio) to recommend biennial screening starting at age 50. Using this threshold, we ascertained reductions of false-positives required to recommend more frequent screening and screening initiation under age 50 using revealed preference analyses.

RESULTS

Using USPSTF implied benefit/harm ratio, routine biennial screening would be recommended starting at 40 if false-positives declined by at least 62%. Reductions of false-positive proportions of 74% would be required to recommend annual screening starting at 40 and reductions of false-positive proportions of 31% would be required to support annual screening starting at 50.

CONCLUSIONS

Using USPSTF revealed preferences, 31-74% reductions in false-positives would be required to recommend mammography screening beyond biennial screening starting at age 50. Widespread implementation of tomosynthesis and reducing recall rates to the lower end of recommended recall rates (5-12%) would provide support for expanding screening beyond biennial screening in women age 50.

Patient, Radiologist, and Examination Characteristics Affecting Screening Mammography Recall Rates in a Large Academic Practice

OBJECTIVE

The aims of this study were to evaluate patient, radiologist, and examination characteristics affecting screening mammography recall rates in an academic breast imaging practice and to identify modifiable factors that could reduce recall variation.

METHODS

This institutional review board-approved retrospective study included screening mammographic examinations in female patients interpreted by 13 breast imaging specialists at an academic center and two outpatient centers from October 1, 2012, to May 31, 2015. Patient demographics were extracted via electronic medical record. Natural language processing captured breast density, BI-RADS assignment, and current and prior screening examination findings. Radiologists' annual screening volumes, clinical experience, and concentration in breast imaging were calculated. Risk aversion, stress from uncertainty, and malpractice concerns were derived via survey. Univariate and multivariate analyses assessed patient, radiologist, and examination characteristics associated with likelihood of mammography recall. The Pearson product-moment correlation coefficient was used to assess the relationship between cancer detection rate and recall rate.

RESULTS

Overall, 5,678 of 61,198 screening examinations (9.3%) were recalled. In multivariate analysis, patient and radiologist characteristics associated with higher odds of recall included patient's age < 50 years (P < .0001), prior mammographic findings (calcification [P < .0001], mass [P < .0001], higher density category [P < .0001]), baseline examination (P < .0001), annual reading volume < 1,250 examinations (P = .0282), and <10 years of experience (P = .0036). Radiologist's risk aversion, stress from uncertainty, malpractice concerns, and cancer detection rates were not associated with higher recall rates (r = -0.36, P = .23).

CONCLUSIONS

In addition to patient and examination factors, screening recall variations were associated with radiologists' annual reading volume and experience. Interventions targeting radiologist factors (screening volumes, second review of potential recalls) may help reduce unwarranted variation in screening recall.