An open letter to panels that are deciding guidelines for breast cancer screening

Evaluating Recalibrating AI Models for Breast Cancer Diagnosis in a New Context: Insights from Transfer Learning, Image Enhancement and High-Quality Training Data Integration

This paper investigates the adaptability of four state-of-the-art artificial intelligence (AI) models to the Australian mammographic context through transfer learning, explores the impact of image enhancement on model performance and analyses the relationship between AI outputs and histopathological features for clinical relevance and accuracy assessment. A total of 1712 screening mammograms (n = 856 cancer cases and n = 856 matched normal cases) were used in this study. The 856 cases with cancer lesions were annotated by two expert radiologists and the level of concordance between their annotations was used to establish two sets: a 'high-concordances subset' with 99% agreement of cancer location and an 'entire dataset' with all cases included. The area under the receiver operating characteristic curve (AUC) was used to evaluate the performance of Globally aware Multiple Instance Classifier (GMIC), Global-Local Activation Maps (GLAM), I&H and End2End AI models, both in the pretrained and transfer learning modes, with and without applying the Contrast Limited Adaptive Histogram Equalization (CLAHE) algorithm. The four AI models with and without transfer learning in the high-concordance subset outperformed those in the entire dataset. Applying the CLAHE algorithm to mammograms improved the performance of the AI models. In the high-concordance subset with the transfer learning and CLAHE algorithm applied, the AUC of the GMIC model was highest (0.912), followed by the GLAM model (0.909), I&H (0.893) and End2End (0.875). There were significant differences (p < 0.05) in the performances of the four AI models between the high-concordance subset and the entire dataset. The AI models demonstrated significant differences in malignancy probability concerning different tumour size categories in mammograms. The performance of AI models was affected by several factors such as concordance classification, image enhancement and transfer learning. Mammograms with a strong concordance with radiologists' annotations, applying image enhancement and transfer learning could enhance the accuracy of AI models.

Multi-view fusion-based local-global dynamic pyramid convolutional cross-tansformer network for density classification in mammography

Object.Breast density is an important indicator of breast cancer risk. However, existing methods for breast density classification do not fully utilise the multi-view information produced by mammography and thus have limited classification accuracy.Method.In this paper, we propose a multi-view fusion network, denoted local-global dynamic pyramidal-convolution transformer network (LG-DPTNet), for breast density classification in mammography. First, for single-view feature extraction, we develop a dynamic pyramid convolutional network to enable the network to adaptively learn global and local features. Second, we address the problem exhibited by traditional multi-view fusion methods, this is based on a cross-transformer that integrates fine-grained information and global contextual information from different views and thereby provides accurate predictions for the network. Finally, we use an asymmetric focal loss function instead of traditional cross-entropy loss during network training to solve the problem of class imbalance in public datasets, thereby further improving the performance of the model.Results.We evaluated the effectiveness of our method on two publicly available mammography datasets, CBIS-DDSM and INbreast, and achieved areas under the curve (AUC) of 96.73% and 91.12%, respectively.Conclusion.Our experiments demonstrated that the devised fusion model can more effectively utilise the information contained in multiple views than existing models and exhibits classification performance that is superior to that of baseline and state-of-the-art methods.

Monitoring Methodology for an AI Tool for Breast Cancer Screening Deployed in Clinical Centers

We propose a methodology for monitoring an artificial intelligence (AI) tool for breast cancer screening when deployed in clinical centers. An AI trained to detect suspicious regions of interest in the four views of a mammogram and to characterize their level of suspicion with a score ranging from one (low suspicion) to ten (high suspicion of malignancy) was deployed in four radiological centers across the US. Results were collected between April 2021 and December 2022, resulting in a dataset of 36,581 AI records. To assess the behavior of the AI, its score distribution in each center was compared to a reference distribution obtained in silico using the Pearson correlation coefficient (PCC) between each center AI score distribution and the reference. The estimated PCCs were 0.998 [min: 0.993, max: 0.999] for center US-1, 0.975 [min: 0.923, max: 0.986] for US-2, 0.995 [min: 0.972, max: 0.998] for US-3 and 0.994 [min: 0.962, max: 0.982] for US-4. These values show that the AI behaved as expected. Low PCC values could be used to trigger an alert, which would facilitate the detection of software malfunctions. This methodology can help create new indicators to improve monitoring of software deployed in hospitals.

Empowering study of breast cancer data with application of artificial intelligence technology: promises, challenges, and use cases

In healthcare, artificial intelligence (AI) technologies have the potential to create significant value by improving time-sensitive outcomes while lowering error rates for each patient. Diagnostic images, clinical notes, and reports are increasingly generated and stored in electronic medical records. This heterogeneous data presenting us with challenges in data analytics and reusability that is by nature has high complexity, thereby necessitating novel ways to store, manage and process, and reuse big data. This presents an urgent need to develop new, scalable, and expandable AI infrastructure and analytical methods that can enable healthcare providers to access knowledge for individual patients, yielding better decisions and outcomes. In this review article, we briefly discuss the nature of data in breast cancer study and the role of AI for generating "smart data" which offer actionable information that supports the better decision for personalized medicine for individual patients. In our view, the biggest challenge is to create a system that makes data robust and smart for healthcare providers and patients that can lead to more effective clinical decision-making, improved health outcomes, and ultimately, managing the healthcare outcomes and costs. We highlight some of the challenges in using breast cancer data and propose the need for an AI-driven environment to address them. We illustrate our vision with practical use cases and discuss a path for empowering the study of breast cancer databases with the application of AI and future directions.

Using BI-RADS Stratifications as Auxiliary Information for Breast Masses Classification in Ultrasound Images

Breast Ultrasound (BUS) imaging has been recognized as an essential imaging modality for breast masses classification in China. Current deep learning (DL) based solutions for BUS classification seek to feed ultrasound (US) images into deep convolutional neural networks (CNNs), to learn a hierarchical combination of features for discriminating malignant and benign masses. One existing problem in current DL-based BUS classification was the lack of spatial and channel-wise features weighting, which inevitably allow interference from redundant features and low sensitivity. In this study, we aim to incorporate the instructive information provided by breast imaging reporting and data system (BI-RADS) within DL-based classification. A novel DL-based BI-RADS Vector-Attention Network (BVA Net) that trains with both texture information and decoded information from BI-RADS stratifications was proposed for the task. Three baseline models, pre-trained DenseNet-121, ResNet-50 and Residual-Attention Network (RA Net) were included for comparison. Experiments were conducted on a large scale private main dataset and two public datasets, UDIAT and BUSI. On the main dataset, BVA Net outperformed other models, in terms of AUC (area under the receiver operating curve, 0.908), ACC (accuracy, 0.865), sensitivity (0.812) and precision (0.795). BVA Net also achieved the high AUC (0.87 and 0.882) and ACC (0.859 and 0.843), on UDIAT and BUSI. Moreover, we proposed a method that integrates both BVA Net binary classification and BI-RADS stratification estimation, called integrated classification. The introduction of integrated classification helped improving the overall sensitivity while maintaining a high specificity.

An interpretable classifier for high-resolution breast cancer screening images utilizing weakly supervised localization

Medical images differ from natural images in significantly higher resolutions and smaller regions of interest. Because of these differences, neural network architectures that work well for natural images might not be applicable to medical image analysis. In this work, we propose a novel neural network model to address these unique properties of medical images. This model first uses a low-capacity, yet memory-efficient, network on the whole image to identify the most informative regions. It then applies another higher-capacity network to collect details from chosen regions. Finally, it employs a fusion module that aggregates global and local information to make a prediction. While existing methods often require lesion segmentation during training, our model is trained with only image-level labels and can generate pixel-level saliency maps indicating possible malignant findings. We apply the model to screening mammography interpretation: predicting the presence or absence of benign and malignant lesions. On the NYU Breast Cancer Screening Dataset, our model outperforms (AUC = 0.93) ResNet-34 and Faster R-CNN in classifying breasts with malignant findings. On the CBIS-DDSM dataset, our model achieves performance (AUC = 0.858) on par with state-of-the-art approaches. Compared to ResNet-34, our model is 4.1x faster for inference while using 78.4% less GPU memory. Furthermore, we demonstrate, in a reader study, that our model surpasses radiologist-level AUC by a margin of 0.11.

Large Scale Semi-Automated Labeling of Routine Free-Text Clinical Records for Deep Learning

Breast cancer is a leading cause of cancer death among women in the USA. Screening mammography is effective in reducing mortality, but has a high rate of unnecessary recalls and biopsies. While deep learning can be applied to mammography, large-scale labeled datasets, which are difficult to obtain, are required. We aim to remove many barriers of dataset development by automatically harvesting data from existing clinical records using a hybrid framework combining traditional NLP and IBM Watson. An expert reviewer manually annotated 3521 breast pathology reports with one of four outcomes: left positive, right positive, bilateral positive, negative. Traditional NLP techniques using seven different machine learning classifiers were compared to IBM Watson's automated natural language classifier. Techniques were evaluated using precision, recall, and F-measure. Logistic regression outperformed all other traditional machine learning classifiers and was used for subsequent comparisons. Both traditional NLP and Watson's NLC performed well for cases under 1024 characters with weighted average F-measures above 0.96 across all classes. Performance of traditional NLP was lower for cases over 1024 characters with an F-measure of 0.83. We demonstrate a hybrid framework using traditional NLP techniques combined with IBM Watson to annotate over 10,000 breast pathology reports for development of a large-scale database to be used for deep learning in mammography. Our work shows that traditional NLP and IBM Watson perform extremely well for cases under 1024 characters and can accelerate the rate of data annotation.

Deep Neural Networks Improve Radiologists' Performance in Breast Cancer Screening

We present a deep convolutional neural network for breast cancer screening exam classification, trained, and evaluated on over 200000 exams (over 1000000 images). Our network achieves an AUC of 0.895 in predicting the presence of cancer in the breast, when tested on the screening population. We attribute the high accuracy to a few technical advances. 1) Our network's novel two-stage architecture and training procedure, which allows us to use a high-capacity patch-level network to learn from pixel-level labels alongside a network learning from macroscopic breast-level labels. 2) A custom ResNet-based network used as a building block of our model, whose balance of depth and width is optimized for high-resolution medical images. 3) Pretraining the network on screening BI-RADS classification, a related task with more noisy labels. 4) Combining multiple input views in an optimal way among a number of possible choices. To validate our model, we conducted a reader study with 14 readers, each reading 720 screening mammogram exams, and show that our model is as accurate as experienced radiologists when presented with the same data. We also show that a hybrid model, averaging the probability of malignancy predicted by a radiologist with a prediction of our neural network, is more accurate than either of the two separately. To further understand our results, we conduct a thorough analysis of our network's performance on different subpopulations of the screening population, the model's design, training procedure, errors, and properties of its internal representations. Our best models are publicly available at https://github.com/nyukat/breast_cancer_classifier.

Male Breast Cancer in the Age of Genetic Testing: An Opportunity for Early Detection, Tailored Therapy, and Surveillance

In detection, treatment, and follow-up, male breast cancer has historically lagged behind female breast cancer. On the whole, breast cancer is less common among men than among women, limiting utility of screening, yet the incidence of male breast cancer is rising, and there are men at high risk for breast cancer. While women at high risk for breast cancer are well characterized, with clearly established guidelines for screening, supplemental screening, risk prevention, counseling, and advocacy, men at high risk for breast cancer are poorly identified and represent a blind spot in public health. Today, more standardized genetic counseling and wider availability of genetic testing are allowing identification of high-risk male relatives of women with breast cancer, as well as men with genetic mutations predisposing to breast cancer. This could provide a new opportunity to update our approach to male breast cancer. This article reviews male breast cancer demographics, risk factors, tumor biology, and oncogenetics; recognizes how male breast cancer differs from its female counterpart; highlights its diagnostic challenges; discusses the implications of the widening clinical use of multigene panel testing; outlines current National Comprehensive Cancer Network guidelines (version 1, 2018) for high-risk men; and explores the possible utility of targeted screening and surveillance. Understanding the current state of male breast cancer management and its challenges is important to shape future considerations for care. Shifting the paradigm of male breast cancer detection toward targeted precision medicine may be the answer to improving clinical outcomes of this uncommon disease. ^©RSNA, 2018.

Online information about risks and benefits of screening mammography in 10 European countries: An observational Web sites analysis

Most publications about breast cancer do not provide accurate and comprehensive information, giving few or no data about risk/benefit ratios. We conducted a comparative study among 10 European countries about health information on breast cancer screening, assessing the first 10 Web sites addressing the general public that appeared following an Internet search.With the help of medical residents involved in the EuroNet MRPH Association, we analyzed the first 30 results of an Internet search in 10 European countries to determine the first 10 sites that offered screening mammography. We searched for the following information: source of information, general information on mammography and breast cancer screening, potential harms and risks (false positives, false positives after biopsy, false negatives, interval cancer, overdiagnosis, lead-time bias, and radiation exposure), and potential benefits (reduced mortality and increased survival).The United Kingdom provided the most information: 39 of all 70 possible identified risks (56%) were reported on its sites. Five nations presented over 35% of the possible information (United Kingdom, Spain, France, Ireland, and Italy); the others were under 30% (Portugal, Poland, Slovenia, Netherlands, and Croatia). Regarding the benefits, sites offering the most complete information were those in France (95%) and Poland (90%).Our results suggest that, despite consensus in the scientific community about providing better information to citizens, further efforts are needed to improve information about breast cancer screening. That conclusion also applies to countries showing better results. We believe that there should be greater coordination in this regard throughout Europe.

Stereotactic Breast Biopsy With Benign Results Does Not Negatively Affect Future Screening Adherence

PURPOSE

To evaluate whether false-positive stereotactic vacuum-assisted breast biopsy (SVAB) affects subsequent mammographic screening adherence.

MATERIALS AND METHODS

This Institutional Review Board-approved, HIPAA-compliant retrospective review of women with SVAB was performed between 2012 and 2014. Patient age, clinical history, biopsy pathology, and first postbiopsy screening mammogram were reviewed. Statistical analyses were performed using Fisher's exact, Mann-Whitney, and χ² tests.

RESULTS

There were 913 SVABs performed in 2012 to 2014 for imaging detected lesions; of these, malignant or high-risk lesions or biopsies resulting in a recommendation of surgical excision were excluded, leaving 395 SVABs yielding benign pathology in 395 women. Findings were matched with a control population consisting of 45,126 women who had a BI-RADS 1 or 2 screening mammogram and did not undergo breast biopsy. In all, 191 of 395 (48.4%) women with a biopsy with benign results and 22,668 of 45,126 (50.2%) women without biopsy returned for annual follow-up >9 months and ≤18 months after the index examination (P = .479). In addition, 57 of 395 (14.4%) women with a biopsy with benign results and 3,336 of 45,126 (7.4%) women without biopsy returned for annual follow-up >18 months after the index examination (P < .001). Older women, women with personal history of breast cancer, and women with postbiopsy complication after benign SVAB were more likely to return for screening (P = .026, P = .028, and P = .026, respectively).

CONCLUSION

The findings in our study suggest that SVABs with benign results do not negatively impact screening mammography adherence. The previously described "harms" of false-positive mammography and biopsy may be exaggerated.

Population-Attributable Risk Proportion of Clinical Risk Factors for Breast Cancer

IMPORTANCE

Many established breast cancer risk factors are used in clinical risk prediction models, although the proportion of breast cancers explained by these factors is unknown.

OBJECTIVE

To determine the population-attributable risk proportion (PARP) for breast cancer associated with clinical breast cancer risk factors among premenopausal and postmenopausal women.

DESIGN, SETTING, AND PARTICIPANTS

Case-control study with 1:10 matching on age, year of risk factor assessment, and Breast Cancer Surveillance Consortium (BCSC) registry. Risk factor data were collected prospectively from January 1, 1996, through October 31, 2012, from BCSC community-based breast imaging facilities. A total of 18 437 women with invasive breast cancer or ductal carcinoma in situ were enrolled as cases and matched to 184 309 women without breast cancer, with a total of 58 146 premenopausal and 144 600 postmenopausal women enrolled in the study.

EXPOSURES

Breast Imaging Reporting and Data System (BI-RADS) breast density (heterogeneously or extremely dense vs scattered fibroglandular densities), first-degree family history of breast cancer, body mass index (>25 vs 18.5-25), history of benign breast biopsy, and nulliparity or age at first birth (≥30 years vs <30 years).

MAIN OUTCOMES AND MEASURES

Population-attributable risk proportion of breast cancer.

RESULTS

Of the 18 437 women with breast cancer, the mean (SD) age was 46.3 (3.7) years among premenopausal women and 61.7 (7.2) years among the postmenopausal women. Overall, 4747 (89.8%) premenopausal and 12 502 (95.1%) postmenopausal women with breast cancer had at least 1 breast cancer risk factor. The combined PARP of all risk factors was 52.7% (95% CI, 49.1%-56.3%) among premenopausal women and 54.7% (95% CI, 46.5%-54.7%) among postmenopausal women. Breast density was the most prevalent risk factor for both premenopausal and postmenopausal women and had the largest effect on the PARP; 39.3% (95% CI, 36.6%-42.0%) of premenopausal and 26.2% (95% CI, 24.4%-28.0%) of postmenopausal breast cancers could potentially be averted if all women with heterogeneously or extremely dense breasts shifted to scattered fibroglandular breast density. Among postmenopausal women, 22.8% (95% CI, 18.3%-27.3%) of breast cancers could potentially be averted if all overweight and obese women attained a body mass index of less than 25.

CONCLUSIONS AND RELEVANCE

Most women with breast cancer have at least 1 breast cancer risk factor routinely documented at the time of mammography, and more than half of premenopausal and postmenopausal breast cancers are explained by these factors. These easily assessed risk factors should be incorporated into risk prediction models to stratify breast cancer risk and promote risk-based screening and targeted prevention efforts.

Investigating young women's motivations to engage in early mammography screening in Switzerland: results of a cross-sectional study

BACKGROUND

The scientific and public debate concerning organized mammography screening is unprecedentedly strong. With research evidence concerning its efficacy being ambiguous, the recommendations pertaining to the age-thresholds for program inclusion vary between - and even within - countries. Data shows that young women who are not yet eligible for systematic screening, have opportunistic mammograms relatively often and, moreover, want to be included in organized programs. Yet, to date, little is known about the precise motivations underlying young women's desire and intentions to go for, not medically indicated, mammographic screening.

METHODS

A cross-sectional online survey was carried out among women aged 30-49 years (n = 918) from Switzerland.

RESULTS

The findings show that high fear (β = .08, p ≤ .05), perceived susceptibility (β = .10, p ≤ .05), and ego-involvement (β = .34, p ≤ .001) are the main predictors of screening intentions among women who are not yet eligible for the systematic program. Also, geographical location (Swiss-French group: β = .15, p ≤ .001; Swiss-Italian group: β = .26, p ≤ .001) and age (β = .11, p ≤ .001) play a role. In turn, breast cancer knowledge, risk perceptions, and educational status do not have a significant impact.

CONCLUSIONS

Young women seem to differ inherently from those who are already eligible for systematic screening in terms of the factors underlying their intentions to engage in mammographic screening. Thus, when striving to promote adherence to systematic screening guidelines - whether based on unequivocal scientific evidence or policy decisions - and to allow women to make evidence-based, informed decisions about mammography, differential strategies are needed to reach different age-groups.

Systematic Analysis and Critical Appraisal of the Quality of the Scientific Evidence and Conflicts of Interest in Practice Guidelines (2005-2013) for Barrett's Esophagus

BACKGROUND

Barrett's esophagus (BE) is a condition that has a small but important risk of progressing to esophageal cancer. To date, no study has assessed the strength of evidence supporting the recommendations for BE. We sought to assess the overall quality of the recommendations and strength of the BE using the AGREE II instrument.

METHODS

A PubMed search was performed to identify guidelines published pertaining to BE. Every guideline was reviewed using the AGREE II format to assess the methodological rigor and validity of the guideline. Additionally, guidelines were reviewed for the level of evidence used to support recommendations, conflicts of interest (COI), and differences in recommendations. Statistical analysis was performed using Stata (version 12).

RESULTS

In total, 234 manuscripts were identified of which 8 guidelines published between 2005 and 2013 pertained to BE. Seventy-five percentage (6/8) graded the evidence used to formulate recommendations. Of the 126 recommendations with supporting evidence, 6 % were supported by level A evidence, 49 % level B evidence, and 45 % level C evidence. Using the AGREE II format, the highest overall assessment grade was the BSG BE guideline (6.5 ± 0.6) followed by the AGA (5.5 ± 0.6). The highest rated domains were scope and purpose (mean 77 range 24-96) and clarity of presentation (mean 75), while the lowest rated domains were editorial independence (mean 32 range 0-92) and applicability of the guideline (mean 35 range 7-90). There was significant variability in recommendations regarding who to screen for BE and surveillance intervals. Finally, only 50 % of the guidelines disclosed if COI were present and 75 % (3/4) reported potentially relevant COI.

CONCLUSIONS

Majority of the BE guideline fail to meet the AGREE II domains, and most of the recommendations are level B or C quality evidence. Further interventions are necessary to improve the overall quality of the guidelines.

Breast cancer screening panels continue to confuse the facts and inject their own biases

Additional confusion has been added to the “debate” about breast cancer. [...]

Breast cancer screening utilization and understanding of current guidelines among rural U.S. women with private insurance

Women living in rural areas of the U.S. face disparities in screening mammography and breast cancer outcomes. We sought to evaluate utilization of mammography, awareness of screening guidelines, and attitudes towards screening among rural insured U.S. women. We conducted a cross-sectional self-administered anonymous survey among 2000 women aged 40-64 insured by the National Rural Electric Cooperative Association, a non-profit insurer for electrical utility workers in predominantly rural areas across the U.S. Outcomes included mammographic screening in the past year, screening interval, awareness of guidelines, and perceived barriers to screening. 1588 women responded to the survey (response rate 79.4 %). 74 % of respondents lived in a rural area. Among women aged 40-49, 66.5 % reported mammographic screening in the past year. 46 % received annual screening, 32 % biennial screening, and 22 % rare/no screening. Among women aged 50-64, 77.1 % reported screening in the past year. 63 % received annual screening, 25 % biennial screening, and 12 % rare/no screening. The majority of women (98 %) believed that the mammography can find breast cancer early and save lives. Less than 1 % of younger women, and only 14 % of women over age 50 identified the recommendations of the U.S. Preventative Services Screening Task Force as the current expert recommendations for screening. Screening practices tended to follow perceived guideline recommendations. When rural U.S. women over age 40 have insurance, most receive breast cancer screening. The screening guidelines of cancer advocacy groups and specialty societies appear more influential and widely recognized than those of the U.S. preventative services taskforce.

Implications of Overdiagnosis: Impact on Screening Mammography Practices

This review article explores the issue of overdiagnosis in screening mammography. Overdiagnosis is the screen detection of a breast cancer, histologically confirmed, that might not otherwise become clinically apparent during the lifetime of the patient. While screening mammography is an imperfect tool, it remains the best tool we have to diagnose breast cancer early, before a patient is symptomatic and at a time when chances of survival and options for treatment are most favorable. In 2015, an estimated 231,840 new cases of breast cancer (excluding ductal carcinoma in situ) will be diagnosed in the United States, and some 40,290 women will die. Despite these data, screening mammography for women ages 40-69 has contributed to a substantial reduction in breast cancer mortality, and organized screening programs have led to a shift from late-stage diagnosis to early-stage detection. Current estimates of overdiagnosis in screening mammography vary widely, from 0% to upwards of 30% of diagnosed cancers. This range reflects the fact that measuring overdiagnosis is not a straightforward calculation, but usually one based on different sets of assumptions and often biased by methodological flaws. The recent development of tomosynthesis, which creates high-resolution, three-dimensional images, has increased breast cancer detection while reducing false recalls. Because the greatest harm of overdiagnosis is overtreatment, the key goal should not be less diagnosis but better treatment decision tools. (Population Health Management 2015;18:S3-S11).