Breast cancer: missed interval and screening-detected cancer at full-field digital mammography and screen-film mammography-- results from a retrospective review

The efficacy of artificial intelligence (AI) in detecting interval cancers in the national screening program of a middle-income country

AIM

We aimed to investigate the efficiency and accuracy of an artificial intelligence (AI) algorithm for detecting interval cancers in a middle-income country's national screening program.

MATERIAL AND METHODS

A total of 2,129,486 mammograms reported as BIRADS 1 and 2 were matched with the national cancer registry for interval cancers (IC). The IC group consisted of 442 cases, of which 36 were excluded due to having mammograms incompatible with the AI system. A control group of 446 women with two negative consequent mammograms was defined as time-proven normal and constituted the normal group. The cancer risk scores of both groups were determined from 1 to 10 with the AI system. The sensitivity and specificity values of the AI system were defined in terms of IC detection. The IC group was divided into subgroups with six-month intervals according to their time from screening to diagnosis: 0-6 months, 6-12 months, 12-18 months, and 18-24 months. The diagnostic performance of the AI system for all patients was evaluated using receiver operating characteristics (ROC) curve analysis. The diagnostic performance of the AI system for major and minor findings that expert readers determined was re-evaluated.

RESULTS

AI labeled 53% of ICs with the highest score of 10. The sensitivity of AI in detecting ICs was 53.7% and 38.5% at specificities of 90% and 95%, respectively. Area under the curve (AUC) of AI in detecting major signs was 0.93 (95% CI: 0.90-0.95) with a sensitivity of 81.6% and 72.4% at specificities of 90% and 95%, respectively (95% CI: 0.73-0.88 and 95% CI: 0.60-0.82 respectively) and minor signs was 0.87 (95% CI: 0.87-0.92) with a sensitivity of 70% and 53% at a specificity of 90% and 95%, respectively (95% CI: 0.65-0.82 and 95% CI: 0.52-0.71 respectively). In subgroup analysis for time to diagnosis, the AUC value of the AI system was higher in the 0-6 month period than in later periods.

CONCLUSION

This study showed the potential of AI in detecting ICs in initial mammograms and reducing human errors and undetected cancers.

Radiological features of screening-detected and interval breast cancers and subsequent survival in Eastern Finnish women

Interval breast cancers are diagnosed between scheduled screenings and differ in many respects from screening-detected cancers. Studies comparing the survival of patients with interval and screening-detected cancers have reported differing results. The aim of this study was to investigate the radiological and histopathological features and growth rates of screening-detected and interval breast cancers and subsequent survival. This retrospective study included 942 female patients aged 50-69 years with breast cancers treated and followed-up at Kuopio University Hospital between January 2010 and December 2016. The screening-detected and interval cancers were classified as true, minimal-signs, missed, or occult. The radiological features were assessed on mammograms by one of two specialist breast radiologists with over 15 years of experience. A χ2 test was used to examine the association between radiological and pathological variables; an unpaired t test was used to compare the growth rates of missed and minimal-signs cancers; and the Kaplan-Meier estimator was used to examine survival after screening-detected and interval cancers. Sixty occult cancers were excluded, so a total of 882 women (mean age 60.4 ± 5.5 years) were included, in whom 581 had screening-detected cancers and 301 interval cancers. Disease-specific survival, overall survival and disease-free survival were all worse after interval cancer than after screening-detected cancer (p < 0.001), with a mean follow-up period of 8.2 years. There were no statistically significant differences in survival between the subgroups of screening-detected or interval cancers. Missed interval cancers had faster growth rates (0.47% ± 0.77%/day) than missed screening-detected cancers (0.21% ± 0.11%/day). Most cancers (77.2%) occurred in low-density breasts (< 25%). The most common lesion types were masses (73.9%) and calcifications (13.4%), whereas distortions (1.8%) and asymmetries (1.7%) were the least common. Survival was worse after interval cancers than after screening-detected cancers, attributed to their more-aggressive histopathological characteristics, more nodal and distant metastases, and faster growth rates.

Imaging characteristics of interval cancers detected on Full Field Digital Mammography (FFDM) versus Digital Breast Tomosynthesis (DBT)

OBJECTIVE

To compare imaging features of interval cancers detected in patients screened with full field digital mammography (FFDM) versus digital breast tomosynthesis (DBT).

MATERIALS/METHODS

This retrospective observational study consisted of female patients undergoing screening DM or FFDM at an academic medical center and two outpatient imaging facilities between January 2012 and June 2017. A natural language processing algorithm queried breast imaging reports for breast density and BI-RADS category. This was cross-referenced to an institutional breast cancer registry to identify interval cancers. Retrospective consensus review of the cases was done to categorize imaging features of interval cancers on FFDM vs DBT.

RESULTS

The rate of interval cancers was comparable in patients screened with FFDM (30/39793) and DBT (29/32180) (p = 0.58). There was no significant difference in the rate, histopathology, or imaging features of interval cancers in patients screened with FFDM versus DBT. The most common mammographic features on diagnostic imaging across both groups was the presence of a mass (13/47). Almost equally common was negative diagnostic mammogram with mass detected only on ultrasound (11/47). The rate of interval cancers detected by high-risk surveillance breast MRI was increased in patients who previously had screening with DBT relative to those who had screening with FFDM (p = 0.0419).

CONCLUSION

There is no significant difference in rate of detection, histopathology, or imaging features of interval cancers in patients screened with FFDM versus DBT. However, across both cohorts, the most common features on diagnostic mammogram were either the presence of a mass or a negative mammogram.

Digital breast tomosynthesis in mammographic screening: false negative cancer cases in the To-Be 1 trial

OBJECTIVES

The randomized controlled trial comparing digital breast tomosynthesis and synthetic 2D mammograms (DBT + SM) versus digital mammography (DM) (the To-Be 1 trial), 2016-2017, did not result in higher cancer detection for DBT + SM. We aimed to determine if negative cases prior to interval and consecutive screen-detected cancers from DBT + SM were due to interpretive error.

METHODS

Five external breast radiologists performed the individual blinded review of 239 screening examinations (90 true negative, 39 false positive, 19 prior to interval cancer, and 91 prior to consecutive screen-detected cancer) and the informed consensus review of examinations prior to interval and screen-detected cancers (n = 110). The reviewers marked suspicious findings with a score of 1-5 (probability of malignancy). A case was false negative if ≥ 2 radiologists assigned the cancer site with a score of ≥ 2 in the blinded review and if the case was assigned as false negative by a consensus in the informed review.

RESULTS

In the informed review, 5.3% of examinations prior to interval cancer and 18.7% prior to consecutive round screen-detected cancer were considered false negative. In the blinded review, 10.6% of examinations prior to interval cancer and 42.9% prior to consecutive round screen-detected cancer were scored ≥ 2. A score of ≥ 2 was assigned to 47.8% of negative and 89.7% of false positive examinations.

CONCLUSIONS

The false negative rates were consistent with those of prior DM reviews, indicating that the lack of higher cancer detection for DBT + SM versus DM in the To-Be 1 trial is complex and not due to interpretive error alone.

CRITICAL RELEVANCE STATEMENT

The randomized controlled trial on digital breast tomosynthesis and synthetic 2D mammograms (DBT) and digital mammography (DM), 2016-2017, showed no difference in cancer detection for the two techniques. The rates of false negative screening examinations prior to interval and consecutive screen-detected cancer for DBT were consistent with the rates in prior DM reviews, indicating that the non-superior DBT performance in the trial might not be due to interpretive error alone.

KEY POINTS

• Screening with digital breast tomosynthesis (DBT) did not result in a higher breast cancer detection rate compared to screening with digital mammography (DM) in the To-Be 1 trial. • The false negative rates for examinations prior to interval and consecutive screen-detected cancer for DBT were determined in the trial to test if the lack of differences was due to interpretive error. • The false negative rates were consistent with those of prior DM reviews, indicating that the lack of higher cancer detection for DBT versus DM was complex and not due to interpretive error alone.

Intratumoral 99mTc-DPD Uptake on Bone Scintigraphy in a Patient With Invasive Micropapillary Breast Carcinoma: A Pathologic Review

ABSTRACT

Invasive micropapillary carcinoma (IMPC) is a rare and aggressive subtype of breast cancer with a poorer prognosis due to high local recurrence and lymphovascular invasion. Interestingly, IMPC often does not show suspicious patterns of calcifications related to malignancy on mammography. Therefore, the lack of suspicious calcifications makes it difficult to detect breast cancer on mammography. With only nonspecific calcifications on mammography, we observed an unusual intratumoral 99mTc-DPD uptake on whole-body bone scintigraphy in an IMPC breast cancer patient during the initial staging workup, and its characteristics were compared with mammographic findings and the postoperative pathologic features.

Breast Cancer Detection with an Ensemble of Deep Learning Networks Using a Consensus-Adaptive Weighting Method

Breast cancer's high mortality rate is often linked to late diagnosis, with mammograms as key but sometimes limited tools in early detection. To enhance diagnostic accuracy and speed, this study introduces a novel computer-aided detection (CAD) ensemble system. This system incorporates advanced deep learning networks-EfficientNet, Xception, MobileNetV2, InceptionV3, and Resnet50-integrated via our innovative consensus-adaptive weighting (CAW) method. This method permits the dynamic adjustment of multiple deep networks, bolstering the system's detection capabilities. Our approach also addresses a major challenge in pixel-level data annotation of faster R-CNNs, highlighted in a prominent previous study. Evaluations on various datasets, including the cropped DDSM (Digital Database for Screening Mammography), DDSM, and INbreast, demonstrated the system's superior performance. In particular, our CAD system showed marked improvement on the cropped DDSM dataset, enhancing detection rates by approximately 1.59% and achieving an accuracy of 95.48%. This innovative system represents a significant advancement in early breast cancer detection, offering the potential for more precise and timely diagnosis, ultimately fostering improved patient outcomes.

Discordant and false-negative interpretations at digital breast tomosynthesis in the prospective Oslo Tomosynthesis Screening Trial (OTST) using independent double reading

OBJECTIVES

To analyze discordant and false-negatives of double reading digital breast tomosynthesis (DBT) versus digital mammography (DM) including reading times in the Oslo Tomosynthesis Screening Trial (OTST), and reclassify these in a retrospective reader study as missed, minimal sign, or true-negatives.

METHODS

The prospective OTST comparing double reading DBT vs. DM had paired design with four parallel arms: DM, DM + computer aided detection, DBT + DM, and DBT + synthetic mammography. Eight radiologists interpreted images in batches using a 5-point scale. Reading time was automatically recorded. A retrospective reader study including four radiologists classified screen-detected cancers with at least one false-negative score and screening examinations of interval cancers as negative, non-specific minimal sign, significant minimal sign, and missed; the two latter groups are defined "actionable." Statistics included chi-square, Fisher's exact, McNemar's, and Mann-Whitney U tests.

RESULTS

Discordant rate (cancer missed by one reader) for screen-detected cancers was overall comparable (DBT (31% [71/227]) and DM (30% [52/175]), p = .81), significantly lower at DBT for spiculated cancers (DBT, 19% [20/106] vs. DM, 36% [38/106], p = .003), but high (28/49 = 57%, p = 0.001) for DBT-only detected spiculated cancers. Reading time and sensitivity varied among readers. False-negative DBT-only detected spiculated cancers had shorter reading time than true-negatives in 46% (13/28). Retrospective evaluation classified the following DBT exams "actionable": three missed by both readers, 95% (39/41) of discordant cancers detected by both modes, all 30 discordant DBT-only cancers, 25% (13/51) of interval cancers.

CONCLUSIONS

Discordant rate was overall comparable for DBT and DM, significantly lower at DBT for spiculated cancers, but high for DBT-only detected spiculated lesions. Most false-negative screen-detected DBT were classified as "actionable."

CLINICAL RELEVANCE STATEMENT

Retrospective evaluation of false-negative interpretations from the Oslo Tomosynthesis Screening Trial shows that most discordant and several interval cancers could have been detected at screening. This underlines the potential for modern AI-based reading aids and triage, as high-volume screening is a demanding task.

KEY POINTS

• Digital breast tomosynthesis (DBT) screening is more sensitive and has higher specificity compared to digital mammography screening, but high-volume DBT screening is a demanding task which can result in high discordance rate among readers. • Independent double reading DBT screening had overall comparable discordance rate as digital mammography, lower for spiculated masses seen on both modalities, and higher for small spiculated cancer seen only on DBT. • Almost all discordant digital breast tomosynthesis-detected cancers (72 of 74) and 25% (13 of 51) of the interval cancers in the Oslo Tomosynthesis Screening Trial were retrospectively classified as actionable and could have been detected by the readers.

Performance of a Breast Cancer Detection AI Algorithm Using the Personal Performance in Mammographic Screening Scheme

Background The Personal Performance in Mammographic Screening (PERFORMS) scheme is used to assess reader performance. Whether this scheme can assess the performance of artificial intelligence (AI) algorithms is unknown. Purpose To compare the performance of human readers and a commercially available AI algorithm interpreting PERFORMS test sets. Materials and Methods In this retrospective study, two PERFORMS test sets, each consisting of 60 challenging cases, were evaluated by human readers between May 2018 and March 2021 and were evaluated by an AI algorithm in 2022. AI considered each breast separately, assigning a suspicion of malignancy score to features detected. Performance was assessed using the highest score per breast. Performance metrics, including sensitivity, specificity, and area under the receiver operating characteristic curve (AUC), were calculated for AI and humans. The study was powered to detect a medium-sized effect (odds ratio, 3.5 or 0.29) for sensitivity. Results A total of 552 human readers interpreted both PERFORMS test sets, consisting of 161 normal breasts, 70 malignant breasts, and nine benign breasts. No difference was observed at the breast level between the AUC for AI and the AUC for human readers (0.93% and 0.88%, respectively; P = .15). When using the developer's suggested recall score threshold, no difference was observed for AI versus human reader sensitivity (84% and 90%, respectively; P = .34), but the specificity of AI was higher (89%) than that of the human readers (76%, P = .003). However, it was not possible to demonstrate equivalence due to the size of the test sets. When using recall thresholds to match mean human reader performance (90% sensitivity, 76% specificity), AI showed no differences inperformance, with a sensitivity of 91% (P =. 73) and a specificity of 77% (P = .85). Conclusion Diagnostic performance of AI was comparable with that of the average human reader when evaluating cases from two enriched test sets from the PERFORMS scheme. © RSNA, 2023 See also the editorial by Philpotts in this issue.

Standalone AI for Breast Cancer Detection at Screening Digital Mammography and Digital Breast Tomosynthesis: A Systematic Review and Meta-Analysis

Background There is considerable interest in the potential use of artificial intelligence (AI) systems in mammographic screening. However, it is essential to critically evaluate the performance of AI before it can become a modality used for independent mammographic interpretation. Purpose To evaluate the reported standalone performances of AI for interpretation of digital mammography and digital breast tomosynthesis (DBT). Materials and Methods A systematic search was conducted in PubMed, Google Scholar, Embase (Ovid), and Web of Science databases for studies published from January 2017 to June 2022. Sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) values were reviewed. Study quality was assessed using the Quality Assessment of Diagnostic Accuracy Studies 2 and Comparative (QUADAS-2 and QUADAS-C, respectively). A random effects meta-analysis and meta-regression analysis were performed for overall studies and for different study types (reader studies vs historic cohort studies) and imaging techniques (digital mammography vs DBT). Results In total, 16 studies that include 1 108 328 examinations in 497 091 women were analyzed (six reader studies, seven historic cohort studies on digital mammography, and four studies on DBT). Pooled AUCs were significantly higher for standalone AI than radiologists in the six reader studies on digital mammography (0.87 vs 0.81, P = .002), but not for historic cohort studies (0.89 vs 0.96, P = .152). Four studies on DBT showed significantly higher AUCs in AI compared with radiologists (0.90 vs 0.79, P < .001). Higher sensitivity and lower specificity were seen for standalone AI compared with radiologists. Conclusion Standalone AI for screening digital mammography performed as well as or better than radiologists. Compared with digital mammography, there is an insufficient number of studies to assess the performance of AI systems in the interpretation of DBT screening examinations. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Scaranelo in this issue.

Impact of age and breast thickness on mean glandular dose of standard digital mammography and digital breast tomosynthesis

This study compares the mean glandular dose (MGD) across 2D, 3D projection and Contrast-Enhanced Digital Mammography (CEDM) mammographic techniques. The important metadata were extracted from the digital mammography console. 650 subjects were clustered based on projections, age and CBT. The MGD of 2D, 3D, and CEDM was positively correlated with CBT but inversely correlated with the age factor. This study indicate MGD of CEDM was 16% and 22% lower compared to 2D and 3D techniques, respectively.

Considerations for Evaluating the Introduction of New Cancer Screening Technology: Use of Interval Cancers to Assess Potential Benefits and Harms

This framework focuses on the importance of the consideration of the downstream intermediate and long-term health outcomes when a change to a screening program is introduced. The authors present a methodology for utilising the relationship between screen-detected and interval cancer rates to infer the benefits and harms associated with a change to the program. A review of the previous use of these measures in the literature is presented. The framework presents other aspects to consider when utilizing this methodology, and builds upon an existing framework that helps researchers, clinicians, and policy makers to consider the impacts of changes to screening programs on health outcomes. It is hoped that this research will inform future evaluative studies to assess the benefits and harms of changes to screening programs.

The Role of Cue-Based Strategies in Skilled Diagnosis Among Pathologists

OBJECTIVE

This research was designed to test whether behavioral indicators of pathology-related cue utilization were associated with performance on a diagnostic task.

BACKGROUND

Across many domains, including pathology, successful diagnosis depends on pattern recognition that is supported by associations in memory in the form of cues. Previous studies have focused on the specific information or knowledge on which medical image expertise relies. The target in this study is the more general ability to identify and interpret relevant information.

METHOD

Data were collected from 54 histopathologists in both conference and online settings. The participants completed a pathology edition of the Expert Intensive Skills Evaluation 2.0 (EXPERTise 2.0) to establish behavioral indicators of context-related cue utilization. They also completed a separate diagnostic task designed to examine related diagnostic skills.

RESULTS

Behavioral indicators of higher or lower cue utilization were based on the participants' performance across five tasks. Accounting for the number of cases reported per year, higher cue utilization was associated with greater accuracy on the diagnostic task. A post hoc analysis suggested that higher cue utilization may be associated with a greater capacity to recognize low prevalence cases.

CONCLUSION

This study provides support for the role of cue utilization in the development and maintenance of skilled diagnosis amongst pathologists.

APPLICATION

Pathologist training needs to be structured to ensure that learners have the opportunity to form cue-based strategies and associations in memory, especially for less commonly seen diseases.

Artificial Intelligence Evaluation of 122 969 Mammography Examinations from a Population-based Screening Program

Background Artificial intelligence (AI) has shown promising results for cancer detection with mammographic screening. However, evidence related to the use of AI in real screening settings remain sparse. Purpose To compare the performance of a commercially available AI system with routine, independent double reading with consensus as performed in a population-based screening program. Furthermore, the histopathologic characteristics of tumors with different AI scores were explored. Materials and Methods In this retrospective study, 122 969 screening examinations from 47 877 women performed at four screening units in BreastScreen Norway from October 2009 to December 2018 were included. The data set included 752 screen-detected cancers (6.1 per 1000 examinations) and 205 interval cancers (1.7 per 1000 examinations). Each examination had an AI score between 1 and 10, where 1 indicated low risk of breast cancer and 10 indicated high risk. Threshold 1, threshold 2, and threshold 3 were used to assess the performance of the AI system as a binary decision tool (selected vs not selected). Threshold 1 was set at an AI score of 10, threshold 2 was set to yield a selection rate similar to the consensus rate (8.8%), and threshold 3 was set to yield a selection rate similar to an average individual radiologist (5.8%). Descriptive statistics were used to summarize screening outcomes. Results A total of 653 of 752 screen-detected cancers (86.8%) and 92 of 205 interval cancers (44.9%) were given a score of 10 by the AI system (threshold 1). Using threshold 3, 80.1% of the screen-detected cancers (602 of 752) and 30.7% of the interval cancers (63 of 205) were selected. Screen-detected cancer with AI scores not selected using the thresholds had favorable histopathologic characteristics compared to those selected; opposite results were observed for interval cancer. Conclusion The proportion of screen-detected cancers not selected by the artificial intelligence (AI) system at the three evaluated thresholds was less than 20%. The overall performance of the AI system was promising according to cancer detection. © RSNA, 2022.

Cancer yield and imaging features of probably benign calcifications at digital magnification view

OBJECTIVES

To investigate the malignancy rate of probably benign calcifications assessed by digital magnification view and imaging and clinical features associated with malignancy.

METHODS

This retrospective study included consecutive women with digital magnification views assessed as probably benign for calcifications without other associated mammographic findings from March 2009 to January 2014. Initial studies rendering a probably benign assessment were analyzed, with biopsy or 4-year imaging follow-up. Fisher's exact test and univariable logistic regression were performed. Cancer yields were calculated.

RESULTS

A total of 458 lesions in 422 patients were finally included. The overall cancer yield was 2.2% (10 of 458, invasive cancer [n = 4] and DCIS [n = 6]). Calcification distribution (OR = 23.80, p = .041), calcification morphology (OR = 10.84, p = .005), increased calcifications (OR = 29.40, p = .001), and having a concurrent newly diagnosed breast cancer or high-risk lesion (OR = 10.24, p = .001) were associated with malignancy. Cancer yields did not significantly differ between grouped punctate calcifications vs. calcifications with other features (1.2% [2 of 162] vs. 2.7% [8 of 296], p = .506). The cancer yield was 1.6% (7 of 437) in women without newly diagnosed breast cancer or high-risk lesions.

CONCLUSION

The cancer yield of probably benign calcifications assessed by digital magnification view was below the 2% threshold for grouped punctate calcifications and for women without newly diagnosed breast cancer or high-risk lesions. Calcification distribution, morphology, increase in calcifications, and the presence of newly diagnosed breast cancer/high-risk lesion were associated with malignancy.

KEY POINTS

• Among 458 probably benign calcifications assessed by digital magnification view, the overall cancer yield was 2.2% (10 of 458). • The cancer yield was below the 2% threshold for grouped punctate calcifications (1.2%, 2 of 162) and in women without newly diagnosed breast cancer or high-risk lesions (1.6%, 7 of 437). • Calcification distribution, morphology, increase in calcifications, and the presence of newly diagnosed breast cancer/high-risk lesion were associated with malignancy (all p < .05).

Retrospective Review of Missed Cancer Detection and Its Mammography Findings with Artificial-Intelligence-Based, Computer-Aided Diagnosis

To investigate whether artificial-intelligence-based, computer-aided diagnosis (AI-CAD) could facilitate the detection of missed cancer on digital mammography, a total of 204 women diagnosed with breast cancer with diagnostic (present) and prior mammograms between 2018 and 2020 were included in this study. Two breast radiologists reviewed the mammographic features and classified them into true negative, minimal sign or missed cancer. They analyzed the AI-CAD results with an abnormality score and assessed whether the AI-CAD correctly localized the known cancer sites. Of the 204 cases, 137 were classified as true negative, 33 as minimal signs, and 34 as missed cancer. The sensitivity, specificity and diagnostic accuracy of AI-CAD were 84.7%, 91.5% and 86.3% on diagnostic mammogram and 67.2%, 91.2% and 83.38% on prior mammogram, respectively. The AI-CAD correctly localized 27 cases from 34 missed cancers on prior mammograms. The findings in the preceding mammography of AI-CAD-detected missed cancer were common in the order of calcifications, focal asymmetry and asymmetry. Asymmetry was the most common finding among the seven cases, which could not be detected by AI-CAD in the missed cases (5/7). The assistance of AI-CAD can be helpful in the early detection of breast cancer in mammography screenings.

High Precision Mammography Lesion Identification From Imprecise Medical Annotations

Breast cancer screening using Mammography serves as the earliest defense against breast cancer, revealing anomalous tissue years before it can be detected through physical screening. Despite the use of high resolution radiography, the presence of densely overlapping patterns challenges the consistency of human-driven diagnosis and drives interest in leveraging state-of-art localization ability of deep convolutional neural networks (DCNN). The growing availability of digitized clinical archives enables the training of deep segmentation models, but training using the most widely available form of coarse hand-drawn annotations works against learning the precise boundary of cancerous tissue in evaluation, while producing results that are more aligned with the annotations rather than the underlying lesions. The expense of collecting high quality pixel-level data in the field of medical science makes this even more difficult. To surmount this fundamental challenge, we propose LatentCADx, a deep learning segmentation model capable of precisely annotating cancer lesions underlying hand-drawn annotations, which we procedurally obtain using joint classification training and a strict segmentation penalty. We demonstrate the capability of LatentCADx on a publicly available dataset of 2,620 Mammogram case files, where LatentCADx obtains classification ROC of 0.97, AP of 0.87, and segmentation AP of 0.75 (IOU = 0.5), giving comparable or better performance than other models. Qualitative and precision evaluation of LatentCADx annotations on validation samples reveals that LatentCADx increases the specificity of segmentations beyond that of existing models trained on hand-drawn annotations, with pixel level specificity reaching a staggering value of 0.90. It also obtains sharp boundary around lesions unlike other methods, reducing the confused pixels in the output by more than 60%.

SCREENING MAMMOGRAPHY: DIAGNOSTIC EFFICACY-ISSUES AND CONSIDERATIONS FOR THE 2020S

Diagnostic efficacy in medical imaging is ultimately a reflection of radiologist performance. This can be influenced by numerous factors, some of which are patient related, such as the physical size and density of the breast, and machine related, where some lesions are difficult to visualise on traditional imaging techniques. Other factors are human reader errors that occur during the diagnostic process, which relate to reader experience and their perceptual and cognitive oversights. Given the large-scale nature of breast cancer screening, even small increases in diagnostic performance equate to large numbers of women saved. It is important to identify the causes of diagnostic errors and how detection efficacy can be improved. This narrative review will therefore explore the various factors that influence mammographic performance and the potential solutions used in an attempt to ameliorate the errors made.

Does it matter for the radiologists' performance whether they read short or long batches in organized mammographic screening?

Objective

To analyze the association between radiologists’ performance and image position within a batch in screen reading of mammograms in Norway.

Method

We described true and false positives and true and false negatives by groups of image positions and batch sizes for 2,937,312 screen readings performed from 2012 to 2018. Mixed-effects models were used to obtain adjusted proportions of true and false positive, true and false negative, sensitivity, and specificity for different image positions. We adjusted for time of day and weekday and included the individual variation between the radiologists as random effects. Time spent reading was included in an additional model to explore a possible mediation effect.

Result

True and false positives were negatively associated with image position within the batch, while the rates of true and false negatives were positively associated. In the adjusted analyses, the rate of true positives was 4.0 per 1000 (95% CI: 3.8–4.2) readings for image position 10 and 3.9 (95% CI: 3.7–4.1) for image position 60. The rate of true negatives was 94.4% (95% CI: 94.0–94.8) for image position 10 and 94.8% (95% CI: 94.4–95.2) for image position 60. Per 1000 readings, the rate of false negative was 0.60 (95% CI: 0.53–0.67) for image position 10 and 0.62 (95% CI: 0.55–0.69) for image position 60.

Conclusion

There was a decrease in the radiologists’ sensitivity throughout the batch, and although this effect was small, our results may be clinically relevant at a population level or when multiplying the differences with the number of screen readings for the individual radiologists.

Key Points

• True and false positive reading scores were negatively associated with image position within a batch.

• A decreasing trend of positive scores indicated a beneficial effect of a certain number of screen readings within a batch.

• False negative scores increased throughout the batch but the association was not statistically significant.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00330-021-08010-9.

Survival among women diagnosed with screen-detected or interval breast cancer classified as true, minimal signs, or missed through an informed radiological review

OBJECTIVES

"True" breast cancers, defined as not being visible on prior screening mammograms, are expected to be more aggressive than "missed" cancers, which are visible in retrospect. However, the evidence to support this hypothesis is limited. We compared the risk of death from any cause for women with true, minimal signs, and missed invasive screen-detected (SDC) and interval breast cancers (IC).

METHODS

This nation-wide study included 1022 SDC and 788 IC diagnosed through BreastScreen Norway during 2005-2016. Cancers were classified as true, minimal signs, or missed by five breast radiologists in a consensus-based informed review of prior screening and diagnostic images. We used multivariable Cox regression to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for the risk of death from any cause associated with true, minimal signs, and missed breast cancers, adjusting for age at diagnosis, histopathologic tumour diameter and grade, and subtype. Separate models were created for SDC and IC.

RESULTS

Among SDC, 463 (44%) were classified as true and 242 (23%) as missed; among IC, 325 (39%) were classified as true and 235 (32%) missed. Missed SDC were associated with a similar risk of death as true SDC (HR = 1.20, 95% CI (0.49, 2.46)). Similar results were observed for missed versus true IC (HR = 1.31, 95% CI (0.77, 2.23)).

CONCLUSIONS

We did not observe a statistical difference in the risk of death for women diagnosed with true or missed SDC or IC; however, the number of cases reviewed and follow-up time limited the precision of our estimates.

KEY POINTS

• An informed radiological review classified screen-detected and interval cancers as true, minimal signs, or missed based on prior screening and diagnostic mammograms. • It has been hypothesised that true cancers, not visible on the prior screening examination, may be more aggressive than missed cancers. • We did not observe a statistical difference in the risk of death from any cause for women with missed versus true screen-detected or interval breast cancers.

Radiological review of prior screening mammograms of screen-detected breast cancer

OBJECTIVE

To perform a radiological review of mammograms from prior screening and diagnosis of screen-detected breast cancer in BreastScreen Norway, a population-based screening program.

METHODS

We performed a consensus-based informed review of mammograms from prior screening and diagnosis for screen-detected breast cancers. Mammographic density and findings on screening and diagnostic mammograms were classified according to the Breast Imaging-Reporting and Data System®. Cases were classified based on visible findings on prior screening mammograms as true (no findings), missed (obvious findings), minimal signs (minor/non-specific findings), or occult (no findings at diagnosis). Histopathologic tumor characteristics were extracted from the Cancer Registry of Norway. The Bonferroni correction was used to adjust for multiple testing; p < 0.001 was considered statistically significant.

RESULTS

The study included mammograms for 1225 women with screen-detected breast cancer. Mean age was 62 years ± 5 (SD); 46% (567/1225) were classified as true, 22% (266/1225) as missed, and 32% (392/1225) as minimal signs. No difference in mammographic density was observed between the classification categories. At diagnosis, 59% (336/567) of true and 70% (185/266) of missed cancers were classified as masses (p = 0.004). The percentage of histological grade 3 cancers was higher for true (30% (138/469)) than for missed (14% (33/234)) cancers (p < 0.001). Estrogen receptor positivity was observed in 86% (387/469) of true and 95% (215/234) of missed (p < 0.001) cancers.

CONCLUSIONS

We classified 22% of the screen-detected cancers as missed based on a review of prior screening mammograms with diagnostic images available. One main goal of the study was quality improvement of radiologists' performance and the program. Visible findings on prior screening mammograms were not necessarily indicative of screening failure.

KEY POINTS

• After a consensus-based informed review, 46% of screen-detected breast cancers were classified as true, 22% as missed, and 32% as minimal signs. • Less favorable prognostic and predictive tumor characteristics were observed in true screen-detected breast cancer compared with missed. • The most frequent mammographic finding for all classification categories at the time of diagnosis was mass, while the most frequent mammographic finding on prior screening mammograms was a mass for missed cancers and asymmetry for minimal signs.

Retrospective radiological review and classification of interval breast cancers within population-based breast screening programmes for the purposes of open disclosure: A systematic review

INTRODUCTION

Interval breast cancers occur following a negative breast screening mammogram and before the next scheduled appointment within screening programmes. Radiological review classifies them as cancers that develop between screens, cancers with no obvious malignant abnormalities on prior screens or cancers not detected at screening. This study aimed to systematically review published literature on the occurrence of open disclosure following interval cancer radiological reviews by breast screening programmes internationally in a retrospective setting and examine methodologies used for radiological reviews for the purposes of disclosure.

METHODS

A search for relevant articles published (January 2000 - May 2019) was conducted according to PICO and PRISMA guidelines. The databases Pubmed, Scopus, Google Scholar, Cinahl, Web of Science, Embase, Science Direct and Global Health were searched. Relevant studies were reviewed if they had completed a retrospective review and classification of interval breast cancers.

RESULTS

Of 46 relevant articles included, no study was identified that conducted a retrospective review purposely for open disclosure. Retrospective reviews were conducted for audit/quality assurance, and research including for radiologist education and learning. Variation in methodology was found across review type (non-blinded/semi-informed approach), number of reviewers and classification categories. The proportion of false negative cancers classified among the studies ranged from 4 to 40 %.

DISCUSSION

Variation among radiological review practices were observed, which likely impacts classification results. To ensure standardised classification of interval breast cancers are employed for the purposes of open disclosure in screening settings, reproducible and consistent methodology is required.

Shotgun proteomics coupled to nanoparticle-based biomarker enrichment reveals a novel panel of extracellular matrix proteins as candidate serum protein biomarkers for early-stage breast cancer detection

Background

The lack of specificity and high degree of false positive and false negative rates when using mammographic screening for detecting early-stage breast cancer is a critical issue. Blood-based molecular assays that could be used in adjunct with mammography for increased specificity and sensitivity could have profound clinical impact. Our objective was to discover and independently verify a panel of candidate blood-based biomarkers that could identify the earliest stages of breast cancer and complement current mammographic screening approaches.

Methods

We used affinity hydrogel nanoparticles coupled with LC-MS/MS analysis to enrich and analyze low-abundance proteins in serum samples from 20 patients with invasive ductal carcinoma (IDC) breast cancer and 20 female control individuals with positive mammograms and benign pathology at biopsy. We compared these results to those obtained from five cohorts of individuals diagnosed with cancer in organs other than breast (ovarian, lung, prostate, and colon cancer, as well as melanoma) to establish IDC-specific protein signatures. Twenty-four IDC candidate biomarkers were then verified by multiple reaction monitoring (LC-MRM) in an independent validation cohort of 60 serum samples specifically including earliest-stage breast cancer and benign controls (19 early-stage (T1a) IDC and 41 controls).

Results

In our discovery set, 56 proteins were increased in the serum samples from IDC patients, and 32 of these proteins were specific to IDC. Verification of a subset of these proteins in an independent cohort of early-stage T1a breast cancer yielded a panel of 4 proteins, ITGA2B (integrin subunit alpha IIb), FLNA (Filamin A), RAP1A (Ras-associated protein-1A), and TLN-1 (Talin-1), which classified breast cancer patients with 100% sensitivity and 85% specificity (AUC of 0.93).

Conclusions

Using a nanoparticle-based protein enrichment technology, we identified and verified a highly specific and sensitive protein signature indicative of early-stage breast cancer with no false positives when assessing benign and inflammatory controls. These markers have been previously reported in cell-ECM interaction and tumor microenvironment biology. Further studies with larger cohorts are needed to evaluate whether this biomarker panel improves the positive predictive value of mammography for breast cancer detection.

Comparison of Mortality Among Participants of Women's Health Initiative Trials With Screening-Detected Breast Cancers vs Interval Breast Cancers

IMPORTANCE

Interval breast cancers (IBCs) are cancers that emerge after a mammogram with negative results but before the patient's next scheduled screening. Interval breast cancer has a worse prognosis than cancers detected by screening; however, it is unknown whether the length of the interscreening period is associated with prognostic features and mortality.

OBJECTIVE

To compare the prognostic features and mortality rate of women with IBCs diagnosed within 1 year or between 1 and 2.5 years of a mammogram with negative results with the prognostic features and mortality rate of women with breast cancers detected by screening.

DESIGN, SETTING, AND PARTICIPANTS

This cohort study used mammography data, tumor characteristics, and patient demographic data from the Women's Health Initiative study, which recruited participants from 1993 to 1998 and followed up with participants for a median of 19 years. The present study sample for these analyses included women aged 50 to 79 years who participated in the Women's Health Initiative study and includes data collected through March 31, 2018. There were 5455 incidents of breast cancer; only 3019 women compliant with screening were retained in analyses. Statistical analysis was performed from October 25, 2018, to November 24, 2019. Breast cancers detected by screening and IBCs were defined based on mammogram history, date of last mammogram, type of visit, and results of examination. Interval breast cancers were subdivided into those occurring within 1 year or between 1 and 2.5 years after the last protocol-mandated mammogram with negative results.

MAIN OUTCOMES AND MEASURES

The primary outcome of this study was breast cancer-specific mortality for each case of breast cancer detected by screening and IBCs detected within 1 year or between 1 and 2.5 years from a mammogram with negative results. Secondary outcomes included prognostic and tumor characteristics for each group. Comparisons between groups were made using the t test, the χ2 test, and Fine-Gray multivariable cumulative incidence regression analyses.

RESULTS

Among the 3019 participants in this analysis, all were women with a mean (SD) age of 63.1 (6.8) years at enrollment and 68.5 (7.1) years at diagnosis. A total of 1050 cases of IBC were identified, with 324 (30.9%) diagnosed within 1 year from a mammogram with negative results and 726 (69.1%) diagnosed between 1 and 2.5 years after last mammogram with negative results. The remaining 1969 cases were breast cancers detected by screening. Interval breast cancers diagnosed within 1 year from a mammogram with negative results had significantly more lobular histologic characteristics (13.0% vs. 8.1%), a larger tumor size (1.97 cm vs 1.43 cm), a higher clinical stage (28.4% vs 17.3% regional and 3.7% vs 0.6% distant), and more lymph node involvement (27.1% vs 17.0%) than cancers detected by screening. Unadjusted breast cancer-specific mortality hazard ratios were significantly higher for IBCs diagnosed within 1 year from a mammogram with negative results compared with breast cancers detected by screening (hazard ratio, 1.92; 95% CI, 1.39-2.65). Higher breast cancer-specific mortality remained statistically significant for IBCs diagnosed within 1 year after adjusting for trial group, molecular subtype, waist to hip ratio, histologic characteristics, and either tumor size (hazard ratio, 1.46; 95% CI, 1.03-2.08) or lymph node involvement (hazard ratio, 1.44; 95% CI, 1.03-2.01). However, significance was lost when tumor size and lymph node involvement were both included in the model (hazard ratio, 1.34; 95% CI, 0.96-1.88). Interval breast cancers diagnosed between 1 and 2.5 years from a mammogram with negative results were not different from breast cancers detected by screening based on prognostic factors or mortality.

CONCLUSIONS AND RELEVANCE

Women with IBCs diagnosed within 1 year of negative mammogram results overall were associated with worse survival than women with breast cancers detected by screening. These differences in survival may be due to a uniquely aggressive biology among IBC cases.

Ultrasound and Clinical Characteristics of False-negative Results in Mammography Screening of Dense Breasts

OBJECTIVE

We analyzed the clinical and ultrasound characteristics associated with false-negative mammography results in women with dense breasts.

MATERIALS AND METHODS

The present study included 191 women (mean age, 54.47 ± 11.61 years; range, 31-75 years) who had presented from July 2015 to June 2018 with pathologically confirmed breast cancer. The mammography, conventional ultrasound, and elastography imaging results of these patients were reviewed. Breast density and screening cancer probability from mammography and conventional ultrasound imaging were scored using the Breast Imaging Reporting and Data System. Multivariate logistic regression analysis was performed to identify the factors independently associated with the false-negative results on breast mammographic screening.

RESULTS

Of 191 confirmed breast cancer cases, 55 (28.8%) were assigned to category ≤ 3, and 136 (71.2%) were assigned to category ≥ 4a according to the mammography findings. All the breasts were graded mammographically as dense. A rougher margin (odds ratio [OR], 8.123; 95% confidence interval [CI], 1.731-38.127) was the strongest independent factor associated with negative results, followed by a lower stiffness ratio (OR, 7.773; 95% CI, 2.574-23.473), negative axillary lymph node status (OR, 5.066; 95% CI, 1.028-24.955), and softer lesions (OR, 1.037; 95% CI, 1.001-1.075).

CONCLUSION

Women with dense breasts, a lower lesion/glandular tissue stiffness ratio, and softer cancer can easily lead to a misdiagnosis using mammography. By giving sufficient attention to the margin, earlier stage cancer with negative lymph node status are more likely to benefit from supplemental ultrasound imaging.

Breast cancers missed during screening in a tertiary-care hospital mammography facility

BACKGROUND

Breast cancer is the most common cancer in females worldwide. Screening with mammography for early breast cancer detection is standard community practice in many countries.

OBJECTIVE

Identify causes of missed breast cancers during screening.

DESIGN

Retrospective, observational.

SETTING

Department of radiology at a tertiary-care hospital mammographic screening facility.

PATIENTS AND METHODS

All women who came with initial negative screens from July 2015 to July 2018 were retrospectively reviewed and followed-up for their second or subsequent mammographic screening. Missed breast cancer was defined as a cancer that was detected on a subsequent mammogram with an initial negative screen. Mammograms were interpreted by two radiologists as per BIRADS (Breast Imaging Reporting and Data System) lexicon. Causes of missed breast cancers were categorized as imaging acquisition (IA), imaging feature (IF) and imaging interpretation (II). True (occult) incident breast cancers were also documented. Percentage estimations for these causes were calculated.

MAIN OUTCOME MEASURES

Breast cancer detection on follow-up screening.

SAMPLE SIZE

943 women.

RESULTS

Of 15 (1.6%) screening-detected breast cancers, 7 cases (46.6%) were missed on the initial screen; 3 (43%) of these were II related, 2 (28.5%) of each were IA and IF. The remaining true (occult) cases were detected on either the second (5 cases) or third screens (3 cases).

CONCLUSION

Improved screening facilities, quality mammographic acquisition and interpretation, double reading, and implementation of an organized screening program may help to avoid missed breast cancers.

LIMITATIONS

Retrospective, small sample, single center, and short duration study.

CONFLICT OF INTEREST

None.

Eye Movements in Medical Image Perception: A Selective Review of Past, Present and Future

The eye movements of experts, reading medical images, have been studied for many years. Unlike topics such as face perception, medical image perception research needs to cope with substantial, qualitative changes in the stimuli under study due to dramatic advances in medical imaging technology. For example, little is known about how radiologists search through 3D volumes of image data because they simply did not exist when earlier eye tracking studies were performed. Moreover, improvements in the affordability and portability of modern eye trackers make other, new studies practical. Here, we review some uses of eye movements in the study of medical image perception with an emphasis on newer work. We ask how basic research on scene perception relates to studies of medical 'scenes' and we discuss how tracking experts' eyes may provide useful insights for medical education and screening efficiency.

[French program of breast cancer screening: Radiologist viewpoint]

French program of breast cancer screening is implemented since15 years and results are in adequation with international guidelines except for participation. To answer to recurrent controversies about breast cancer screening, publications from National Institute of French cancer registry confirm the positive impact of screening on decreasing mortality for participating women. The harms of mammography (and not from screening) need to be communicated to the invited women to help them to make decision about participation but also the risk of worse prognosis in case of symptomatic cancer. The future of screening will be different and works are in progress to find new ways to select women who will beneficiate for screening and whose cancer needs to be treated. Until then, the only way to screen for breast cancer stays the mammographic process as well as other technics in case of dense breast or in case of family history of breast cancer.

Artificial intelligence in breast imaging

This article reviews current limitations and future opportunities for the application of computer-aided detection (CAD) systems and artificial intelligence in breast imaging. Traditional CAD systems in mammography screening have followed a rules-based approach, incorporating domain knowledge into hand-crafted features before using classical machine learning techniques as a classifier. The first commercial CAD system, ImageChecker M1000, relies on computer vision techniques for pattern recognition. Unfortunately, CAD systems have been shown to adversely affect some radiologists' performance and increase recall rates. The Digital Mammography DREAM Challenge was a multidisciplinary collaboration that provided 640,000 mammography images for teams to help decrease false-positive rates in breast cancer screening. Winning solutions leveraged deep learning's (DL) automatic hierarchical feature learning capabilities and used convolutional neural networks. Start-ups Therapixel and Kheiron Medical Technologies are using DL for breast cancer screening. With increasing use of digital breast tomosynthesis, specific artificial intelligence (AI)-CAD systems are emerging to include iCAD's PowerLook Tomo Detection and ScreenPoint Medical's Transpara. Other AI-CAD systems are focusing on breast diagnostic techniques such as ultrasound and magnetic resonance imaging (MRI). There is a gap in the market for contrast-enhanced spectral mammography AI-CAD tools. Clinical implementation of AI-CAD tools requires testing in scenarios mimicking real life to prove its usefulness in the clinical environment. This requires a large and representative dataset for testing and assessment of the reader's interaction with the tools. A cost-effectiveness assessment should be undertaken, with a large feasibility study carried out to ensure there are no unintended consequences. AI-CAD systems should incorporate explainable AI in accordance with the European Union General Data Protection Regulation (GDPR).

Impact of full field digital mammography diagnosis for female patients with breast cancer

BACKGROUND

Previous clinical studies have reported that full field digital mammography (FFDM) can be used for diagnosis on breast cancer (BC) with promising outcome results. However, no study systematically investigates its diagnostic impact on female patients with BC. Thus, this systematic review will assess the accurate of FFDM diagnosis on BC.

METHODS

In this study, we will perform a comprehensive search strategy in the databases as follows: Cochrane Library, EMBASE, MEDILINE, PSYCINFO, Web of Science, Cumulative Index to Nursing and Allied Health Literature, Allied and Complementary Medicine Database, Chinese Biomedical Literature Database, China National Knowledge Infrastructure, VIP Information, and Wanfang Data from inception to February 28, 2019. All case-controlled studies exploring the impacts of FFDM diagnosis for patients BC will be fully considered for inclusion in this study. Two authors will independently scan the title and abstracts for relevance, and assess full texts for inclusion. They will also independently extract data and will assess methodological qualify for each included study by using Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool. RevMan V.5.3 software (London, UK) and Stata V.12.0 software (Texas, USA) will be used to pool the data and to conduct the meta-analysis.

RESULTS

The sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, and diagnostic odds ratio of FFDM will be used to determine the diagnostic accuracy of FFDM for the diagnosis of patients with BC.

CONCLUSION

Its findings will provide latest evidence for the diagnostic accuracy of FFDM in female patients with BC.

SYSTEMATIC REVIEW REGISTRATION

PROSPERO CRD42019125338.

Can high school students help to improve breast radiologists in detecting missed breast cancer lesions on full-field digital mammography?

Aim: To investigate whether full-field digital mammography (FFDM) and contrast-enhanced mammography (CEM), evaluated by non-experienced high school students, improves detection of missed breast cancer lesions on FFDM, in the same cohort of patients. Methods: Non-experienced first- and second year high school students examined fourteen cases of patients diagnosed with breast cancer. These cases consisted of missed breast cancer lesions on FFDM by a breast radiologist. Sensitivity of assessment of the students on FFDM and CEM was analysed and compared with the initial results of the breast radiologists. Results: A total of 134 high school students participated in this study. Mean age was 12.8 years (range 10-14). Based on FFDM, mean overall sensitivity of the students was 29.2% (18.9 - 39.6%). When recombined CEM images were used, mean overall sensitivity of students improved to 82.6% (74.0 - 91.2%) (p=0.001). Mean overall sensitivity of FFDM exams evaluated by radiologists was 75.7% (64.2 - 87.3%), which was lower when compared to student's evaluations on recombined CEM exams, yet not statistically significant (p=0.098). Conclusions: Contrast-enhanced mammography evaluated by non-experienced high school students might improve detection rate of breast cancer when compared to evaluations of only full-field digital mammography by radiologists.

Screening mammography for second breast cancers in women with history of early-stage breast cancer: factors and causes associated with non-detection

Background

The aim of our study was to identify the factors and causes associated with non-detection for second breast cancers on screening mammography in women with a personal history of early-stage breast cancer.

Methods

Between January 2000 and December 2008, 7976 women with early-stage breast cancer underwent breast surgery in our institution. The inclusion criteria of our study were patients who had: (a) subsequent in-breast recurrence, (b) surveillance mammography within 1 year before recurrence. Retrospective analysis of mammography was performed. Non-detection was defined as second breast cancers that were not visible on screening mammography. Imaging features, demographics, primary breast cancer (PBC) characteristics, and clinical features were evaluated to determine its association with non-detection. Univariate and multivariate logistic regression analyses were also performed to identify the factors related to non-detection.

Results

We identified 188 patients that met the criteria. Among them, 39% of patients showed non-detection (n = 74). Of the 74 patients with non-detection, 53 (72%) were classified as having no detectable mammographic abnormality (i.e., true negative) due to overlapping dense breast tissue (n = 32), obscured by postoperative scar (n = 12) or difficult anatomic location / poor positioning (n = 9). The remaining 21 patients were categorized as having subtle findings (n = 11) or missed cancer (n = 10). Non-detection for second breast cancers were significantly associated with mammographic breast density (p = 0.001, OR = 2.959) and detectability of PBC on mammography (p = 0.011, OR = 3.013).

Conclusion

Non-detection of second breast cancer in women with a personal history of early-stage breast cancer were associated with mammographic dense breast and lower detectability of PBC on mammography.

Detecting and classifying lesions in mammograms with Deep Learning

In the last two decades, Computer Aided Detection (CAD) systems were developed to help radiologists analyse screening mammograms, however benefits of current CAD technologies appear to be contradictory, therefore they should be improved to be ultimately considered useful. Since 2012, deep convolutional neural networks (CNN) have been a tremendous success in image recognition, reaching human performance. These methods have greatly surpassed the traditional approaches, which are similar to currently used CAD solutions. Deep CNN-s have the potential to revolutionize medical image analysis. We propose a CAD system based on one of the most successful object detection frameworks, Faster R-CNN. The system detects and classifies malignant or benign lesions on a mammogram without any human intervention. The proposed method sets the state of the art classification performance on the public INbreast database, AUC = 0.95. The approach described here has achieved 2nd place in the Digital Mammography DREAM Challenge with AUC = 0.85. When used as a detector, the system reaches high sensitivity with very few false positive marks per image on the INbreast dataset. Source code, the trained model and an OsiriX plugin are published online at https://github.com/riblidezso/frcnn_cad .

Detection and interval cancer rates during the transition from screen-film to digital mammography in population-based screening

Background

Between 2003 and 2010 digital mammography (DM) gradually replaced screen-film mammography (SFM) in the Dutch breast cancer screening programme (BCSP). Previous studies showed increases in detection rate (DR) after the transition to DM. However, national interval cancer rates (ICR) have not yet been reported.

Methods

We assessed programme sensitivity and specificity during the transition period to DM, analysing nationwide data on screen-detected and interval cancers. Data of 7.3 million screens in women aged 49–74, between 2004 and 2011, were linked to the Netherlands Cancer Registry to obtain data on interval cancers. Age-adjusted DRs, ICRs and recall rates (RR) per 1000 screens and programme sensitivity and specificity were calculated by year, age and screening modality.

Results

41,662 screen-detected and 16,160 interval cancers were analysed. The DR significantly increased from 5.13 (95% confidence interval (CI):5.00–5.30) in 2004 to 6.34 (95% CI:6.15–6.47) in 2011, for both in situ (2004:0.73;2011:1.24) and invasive cancers (2004:4.42;2011:5.07), whereas the ICR remained stable (2004: 2.16 (95% CI2.06–2.25);2011: 2.13 (95% CI:2.04–2.22)). The RR changed significantly from 14.0 to 21.4. Programme sensitivity significantly increased, mainly between ages 49–59, from 70.0% (95% CI:68.9–71.2) to 74.4% (95% CI:73.5–75.4) whereas specificity slightly declined (2004:99.1% (95% CI:99.09–99.13);2011:98.5% (95% CI:98.45–98.50)). The overall DR was significantly higher for DM than for SFM (6.24;5.36) as was programme sensitivity (73.6%;70.1%), the ICR was similar (2.19;2.20) and specificity was significantly lower for DM (98.5%;98.9%).

Conclusions

During the transition from SFM to DM, there was a significant rise in DR and a stable ICR, leading to increased programme sensitivity. Although the recall rate increased, programme specificity remained high compared to other countries. These findings indicate that the performance of DM in a nationwide screening programme is not inferior to, and may be even better, than that of SFM.

Electronic supplementary material

The online version of this article (10.1186/s12885-018-4122-2) contains supplementary material, which is available to authorized users.

Errors in Mammography Cannot be Solved Through Technology Alone

Mammography has been the frontline screening tool for breast cancer for decades. However, high error rates in the form of false negatives (FNs) and false positives (FPs) have persisted despite technological improvements. Radiologists still miss between 10% and 30% of cancers while 80% of woman recalled for additional views have normal outcomes, with 40% of biopsied lesions being benign. Research show that the majority of cancers missed is actually visible and looked at, but either go unnoticed or are deemed to be benign. Causal agents for these errors include human related characteristics resulting in contributory search, perception and decision-making behaviours. Technical, patient and lesion factors are also important relating to positioning, compression, patient size, breast density and presence of breast implants as well as the nature and subtype of the cancer itself, where features such as architectural distortion and triple-negative cancers remain challenging to detect on screening. A better understanding of these causal agents as well as the adoption of technological and educational interventions, which audits reader performance and provide immediate perceptual feedback, should help. This paper reviews the current status of our knowledge around error rates in mammography and explores the factors impacting it. It also presents potential solutions for maximizing diagnostic efficacy thus benefiting the millions of women who undergo this procedure each year.

Incidence and tumour characteristics of bilateral and unilateral interval breast cancers at screening mammography

BACKGROUND

Detected by screening mammography, bilateral breast cancer has a different pathological profile compared to unilateral breast cancer. We investigated the incidence of bilateral interval breast cancers and compared their characteristics with those of unilateral interval breast cancers.

METHODS

We included all 468,720 screening mammograms of women who underwent biennial screening mammography in the South of the Netherlands between January 2005 and January 2015. We collected breast imaging reports, biopsy results and surgical reports of all referred women and of all women who presented with interval breast cancer. The tumour with the highest tumour stage (index cancer) was used for comparison with unilateral interval cancers.

RESULTS

A total of 753 interval cancers were detected, of which 24 (3.2%) were bilateral. Among the invasive interval cancers, bilateral cancers more frequently showed a lobular histology than unilateral cancers (37.5% (9/24) vs. 16.1% (111/691), P = .01). There is a trend towards a larger proportion of bilateral than unilateral interval cancers graded 1 (45.8% (11/24) vs. 27.8% (192/691), P = .08). There were no other statistically significant differences in tumour characteristics. Also, the proportion of interval cancers showing significant mammographic abnormalities at the latest screen was comparable for unilateral and bilateral interval cancers (23.0% vs. 25.0%, P = .9).

DISCUSSION

Bilateral interval cancers comprise a small proportion of all interval cancers. Except of a higher proportion of invasive lobular cancers and a more favourable histological grade of invasive cancers, tumour characteristics are comparable for bilateral and unilateral interval breast cancers.

Effectiveness of and overdiagnosis from mammography screening in the Netherlands: population based study

Objective To analyse stage specific incidence of breast cancer in the Netherlands where women have been invited to biennial mammography screening since 1989 (ages 50-69) and 1997 (ages 70-75), and to assess changes in breast cancer mortality and quantified overdiagnosis.Design Population based study.Setting Mammography screening programme, the Netherlands.Participants Dutch women of all ages, 1989 to 2012.Main outcome measures Stage specific age adjusted incidence of breast cancer from 1989 to 2012. The extra numbers of in situ and stage 1 breast tumours associated with screening were estimated by comparing rates in women aged 50-74 with those in age groups not invited to screening. Overdiagnosis was estimated after subtraction of the lead time cancers. Breast cancer mortality reductions and overdiagnosis during 2010-12 were computed without (scenario 1) and with (scenario 2) a cohort effect on mortality secular trends.Results The incidence of stage 2-4 breast cancers in women aged 50 or more was 168 per 100 000 in 1989 and 166 per 100 000 in 2012. Screening would be associated with a 5% mortality reduction in scenario 1 and with no influence on mortality in scenario 2. In both scenarios, improved treatments would be associated with 28% reductions in mortality. Overdiagnosis has steadily increased over time with the extension of screening to women aged 70-75 and with the introduction of digital mammography. After deduction of clinical lead time cancers, 33% of cancers found in women invited to screening in 2010-12 and 59% of screen detected cancers would be overdiagnosed.Conclusions The Dutch mammography screening programme seems to have little impact on the burden of advanced breast cancers, which suggests a marginal effect on breast cancer mortality. About half of screen detected breast cancers would represent overdiagnosis.

Comparing search patterns in digital breast tomosynthesis and full-field digital mammography: an eye tracking study

As a promising imaging modality, digital breast tomosynthesis (DBT) leads to better diagnostic performance than traditional full-field digital mammograms (FFDM) alone. DBT allows different planes of the breast to be visualized, reducing occlusion from overlapping tissue. Although DBT is gaining popularity, best practices for search strategies in this medium are unclear. Eye tracking allowed us to describe search patterns adopted by radiologists searching DBT and FFDM images. Eleven radiologists examined eight DBT and FFDM cases. Observers marked suspicious masses with mouse clicks. Eye position was recorded at 1000 Hz and was coregistered with slice/depth plane as the radiologist scrolled through the DBT images, allowing a 3-D representation of eye position. Hit rate for masses was higher for tomography cases than 2-D cases and DBT led to lower false positive rates. However, search duration was much longer for DBT cases than FFDM. DBT was associated with longer fixations but similar saccadic amplitude compared with FFDM. When comparing radiologists' eye movements to a previous study, which tracked eye movements as radiologists read chest CT, we found DBT viewers did not align with previously identified "driller" or "scanner" strategies, although their search strategy most closely aligns with a type of vigorous drilling strategy.

Developing asymmetry in a screening mammogram: A cautionary tale of a missed cancer

The developing asymmetry has a 12-15% risk of malignancy but poses challenges of detection and interpretation due to the lack of typical features of cancer and the frequent absence of an ultrasound correlate. Failure to biopsy these lesions may lead to delayed diagnosis of breast cancer.

Carcinoma in situ of the breast in New South Wales, Australia: Current status and trends over the last 40 year

BACKGROUND

The incidence of non-invasive breast cancer has increased substantially over time. We aim to describe temporal trends in the incidence of carcinoma in situ of the breast in New South Wales (NSW), Australia.

METHODS

Descriptive study of trends in the incidence of ductal carcinoma in situ (DCIS) and lobular carcinoma in situ (LCIS) in women who received a diagnosis from 1972 to 2012, recorded in the NSW Cancer Registry.

RESULTS

Carcinoma in situ as a proportion of all breast cancer was 0.4% during the prescreening period 1972 to 1987 and is currently 14.1% (2006 to 2012). Among 10,810 women diagnosed with DCIS, incidence across all ages rose from 0.15 per 100,000 during 1972 to 1983 to 16.81 per 100,000 over 2006 to 2012, representing a 100-fold increase (IRR 113.10; 95% CI 81.94 to 156.08). Among women in the target age group for screening (50-69 years) incidence rose from 0.27 per 100,000 to 51.96 over the same period (IRR 195.50; 95% CI 117.26 to 325.89). DCIS incidence peaks in women aged 60-69 years. DCIS incidence has not stabilized despite screening being well established for over 20 years, and participation rates in the target age range remaining stable.

CONCLUSIONS

Our findings raise questions about the value of the increasing detection of DCIS and aggressive treatment of these lesions, especially among older women, and support trials of de-escalated treatment.

Interval breast cancer characteristics before, during and after the transition from screen-film to full-field digital screening mammography

Background

To determine the proportion of “true” interval cancers and tumor characteristics of interval breast cancers prior to, during and after the transition from screen-film mammography screening (SFM) to full-field digital mammography screening (FFDM).

Methods

We included all women with interval cancers detected between January 2006 and January 2014. Breast imaging reports, biopsy results and breast surgery reports of all women recalled at screening mammography and of all women with interval breast cancers were collected. Two experienced screening radiologists reviewed the diagnostic mammograms, on which the interval cancers were diagnosed, as well as the prior screening mammograms and determined whether or not the interval cancer had been missed on the most recent screening mammogram. If not missed, the cancer was considered an occult (“true”) interval cancer.

Results

A total of 442 interval cancers had been diagnosed, of which 144 at SFM with a prior SFM (SFM-SFM), 159 at FFDM with a prior SFM (FFDM-SFM) and 139 at FFDM with a prior FFDM (FFDM-FFDM). The transition from SFM to FFDM screening resulted in the diagnosis of more occult (“true”) interval cancers at FFDM-SFM than at SFM-SFM (65.4% (104/159) versus 49.3% (71/144), P < 0.01), but this increase was no longer statistically significant in women who had been screened digitally for the second time (57.6% (80/139) at FFDM-FFDM versus 49.3% (71/144) at SFM-SFM). Tumor characteristics were comparable for the three interval cancer cohorts, except of a lower porportion (75.7 and 78.0% versus 67.2% af FFDM-FFDM, P < 0.05) of invasive ductal cancers at FFDM with prior FFDM.

Conclusions

An increase in the proportion of occult interval cancers is observed during the transition from SFM to FFDM screening mammography. However, this increase seems temporary and is no longer detectable after the second round of digital screening. Tumor characteristics and type of surgery are comparable for interval cancers detected prior to, during and after the transition from SFM to FFDM screening mammography, except of a lower proportion of invasive ductal cancers after the transition.

Impact of the Introduction of Digital Mammography in an Organized Screening Program on the Recall and Detection Rate

In 2012, the Reggio Emilia Breast Cancer Screening Program introduced digital mammography in all its facilities at the same time. The aim of this work is to analyze the impact of digital mammography introduction on the recall rate, detection rate, and positive predictive value. The program actively invites women aged 45-74 years. We included women screened in 2011, all of whom underwent film-screen mammography, and all women screened in 2012, all of whom underwent digital mammography. Double reading was used for all mammograms, with arbitration in the event of disagreement. A total of 42,240 women underwent screen-film mammography and 45,196 underwent digital mammography. The recall rate increased from 3.3 to 4.4% in the first year of digital mammography (relative recall adjusted by age and round 1.46, 95% CI = 1.37-1.56); the positivity rate for each individual reading, before arbitration, rose from 3 to 5.7%. The digital mammography recall rate decreased during 2012: after 12 months, it was similar to the recall rate with screen-film mammography. The detection rate was similar: 5.9/1000 and 5.2/1000 with screen-film and digital mammography, respectively (adjusted relative detection rate 0.95, 95% CI = 0.79-1.13). The relative detection rate for ductal carcinoma in situ remained the same. The introduction of digital mammography to our organized screening program had a negative impact on specificity, thereby increasing the recall rate. The effect was limited to the first 12 months after introduction and was attenuated by the double reading with arbitration. We did not observe any relevant effects on the detection rate.

Use-inspired basic research in medical image perception

This journal is dedicated to "use-inspired basic research" where a problem in the world shapes the hypotheses for a study in the laboratory. This brief review presents several examples of "use-inspired basic research" in the area of medical image perception. These are cases where the field of radiology raises an interesting issue in visual cognition. Basic research on those issues may then lead to proposals to improve performance on clinical tasks in medical image perception. Of the six examples given here, the first three ask essentially perceptual questions: How can stereopsis improve medical image perception? How shall we assess the tradeoff between radiation dose and image quality? How does the choice of colors change the interpretation of medical images? The second three examples address attentional issues in those aspects of radiology that can be described as visual search problems: Can eye tracking help us understand errors in radiologic search? What happens if the number of targets in an image is unknown? What happens if, as in radiology screening programs, the target of search is very rare?