Mammographic Interpretation Research Articles

A systematic scoping review exploring variation in practice in specimen mammography for Intraoperative Margin Analysis in Breast Conserving Surgery and the role of artificial intelligence in optimising diagnostic accuracy

PurposeSpecimen Mammography (SM) is commonly used in Breast Conserving Surgery (BCS) for intraoperative margin analysis. A systematic scoping review was conducted to identify sources of methodological variation in Specimen Mammography Interpretation (SMI) and assess the role of Artificial Intelligence (AI) techniques to optimise Diagnostic Accuracy (DA). MethodsEmbase, Pubmed, Cochrane and web of science databases were searched. Studies were included if SM was used for margin analysis for BCS with reported DA compared with pathological margin status and data extracted. Results1242 unique studies were identified, of which 40 were included. 39/40 studies did not utilise AI for SMI, with 4 studies comparing 2 relevant techniques, giving 43 non-AI study arms for analysis. There was wide variation in SM techniques, including number of views and location of SM. Specialist performing SMI in usual clinical practice was surgeon (13/39 studies;33 %), radiologist(s) (16/39;41 %), surgeon and radiologist (3/39;8 %) or not stated (7/39;18 %) which differed from the study specialist in 15/39 (38 %) of studies. Diagnostic accuracy in studies ranged from sensitivity 19–91.7 % and specificity 25–100 %. ConclusionsThere is marked variation in current techniques used for SM for intraoperative margin analysis with correspondingly disparate DA. Only 1 study applied AI to SMI, and we identify how AI could optimise SMI and a template for future work to apply AI techniques to SMI, reduce unwarranted variation and optimise DA.

Optimizing mammography interpretation education: leveraging deep learning for cohort-specific error detection to enhance radiologist training.

Accurate interpretation of mammograms presents challenges. Tailoring mammography training to reader profiles holds the promise of an effective strategy to reduce these errors. This proof-of-concept study investigated the feasibility of employing convolutional neural networks (CNNs) with transfer learning to categorize regions associated with false-positive (FP) errors within screening mammograms into categories of "low" or "high" likelihood of being a false-positive detection for radiologists sharing similar geographic characteristics. Mammography test sets assessed by two geographically distant cohorts of radiologists (cohorts A and B) were collected. FP patches within these mammograms were segmented and categorized as "difficult" or "easy" based on the number of readers committing FP errors. Patches outside 1.5 times the interquartile range above the upper quartile were labeled as difficult, whereas the remaining patches were labeled as easy. Using transfer learning, a patch-wise CNN model for binary patch classification was developed utilizing ResNet as the feature extractor, with modified fully connected layers for the target task. Model performance was assessed using 10-fold cross-validation. Compared with other architectures, the transferred ResNet-50 achieved the highest performance, obtaining receiver operating characteristics area under the curve values of 0.933 ( ) and 0.975 ( ) on the validation sets for cohorts A and B, respectively. The findings highlight the feasibility of employing CNN-based transfer learning to predict the difficulty levels of local FP patches in screening mammograms for specific radiologist cohort with similar geographic characteristics.

Mammographic features differ with body composition in women with breast cancer.

There are several breast cancer (BC) risk factors-many related to body composition, hormonal status, and fertility patterns. However, it is not known if risk factors in healthy women are associated with specific mammographic features at the time of BC diagnosis. Our aim was to assess the potential association between pre-diagnostic body composition and mammographic features in the diagnostic BC image. The prospective Malmö Diet and Cancer Study includes women with invasive BC from 1991 to 2014 (n = 1116). BC risk factors at baseline were registered (anthropometric measures, menopausal status, and parity) along with mammography data from BC diagnosis (breast density, mammographic tumor appearance, and mode of detection). We investigated associations between anthropometric measures and mammographic features via logistic regression analyses, yielding odds ratios (OR) with 95% confidence intervals (CI). There was an association between high body mass index (BMI) (≥ 30) at baseline and spiculated tumor appearance (OR 1.370 (95% CI: 0.941-2.010)), primarily in women with clinically detected cancers (OR 2.240 (95% CI: 1.280-3.940)), and in postmenopausal women (OR 1.580 (95% CI: 1.030-2.440)). Furthermore, an inverse association between high BMI (≥ 30) and high breast density (OR 0.270 (95% CI: 0.166-0.438)) was found. This study demonstrated an association between obesity and a spiculated mass on mammography-especially in women with clinically detected cancers and in postmenopausal women. These findings offer insights on the relationship between risk factors in healthy women and related mammographic features in subsequent BC. With increasing numbers of both BC incidence and women with obesity, it is important to highlight mammographic findings in women with an unhealthy weight. Women with obesity and BC may present with certain mammographic features. Spiculated masses were more common in women with obesity, especially postmenopausal women, and those with clinically detected BCs. Insights on the relationship between obesity and related mammographic features will aid mammographic interpretation.

Screening Outcomes of Mammography with AI in Dense Breasts: A Comparative Study with Supplemental Screening US.

Background Comparative performance between artificial intelligence (AI) and breast US for women with dense breasts undergoing screening mammography remains unclear. Purpose To compare the performance of mammography alone, mammography with AI, and mammography plus supplemental US for screening women with dense breasts, and to investigate the characteristics of the detected cancers. Materials and Methods A retrospective database search identified consecutive asymptomatic women (≥40 years of age) with dense breasts who underwent mammography plus supplemental whole-breast handheld US from January 2017 to December 2018 at a primary health care center. Sequential reading for mammography alone and mammography with the aid of an AI system was conducted by five breast radiologists, and their recall decisions were recorded. Results of the combined mammography and US examinations were collected from the database. A dedicated breast radiologist reviewed marks for mammography alone or with AI to confirm lesion identification. The reference standard was histologic examination and 1-year follow-up data. The cancer detection rate (CDR) per 1000 screening examinations, sensitivity, specificity, and abnormal interpretation rate (AIR) of mammography alone, mammography with AI, and mammography plus US were compared. Results Among 5707 asymptomatic women (mean age, 52.4 years ± 7.9 [SD]), 33 (0.6%) had cancer (median lesion size, 0.7 cm). Mammography with AI had a higher specificity (95.3% [95% CI: 94.7, 95.8], P = .003) and lower AIR (5.0% [95% CI: 4.5, 5.6], P = .004) than mammography alone (94.3% [95% CI: 93.6, 94.8] and 6.0% [95% CI: 5.4, 6.7], respectively). Mammography plus US had a higher CDR (5.6 vs 3.5 per 1000 examinations, P = .002) and sensitivity (97.0% vs 60.6%, P = .002) but lower specificity (77.6% vs 95.3%, P < .001) and higher AIR (22.9% vs 5.0%, P < .001) than mammography with AI. Supplemental US alone helped detect 12 cancers, mostly stage 0 and I (92%, 11 of 12). Conclusion Although AI improved the specificity of mammography interpretation, mammography plus supplemental US helped detect more node-negative early breast cancers that were undetected using mammography with AI. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Whitman and Destounis in this issue.

Parallel Hierarchical Multi-View Feature Fusion Based on Canonical Correlation Analysis for Mammogram Retrieval

Due to the diversity of image sources, content-based multi-source image fusion and retrieval have shown promising capabilities in computer vision tasks, and especially when applied in Computer-Aided Diagnosis (CAD) to automate and improve the accuracy of medical image analysis. The combination of computer vision and CAD systems has the potential to revolutionize healthcare by augmenting the expertise of clinicians, improving overall diagnostic accuracy and helping experts in the clinical decision-making process by classifying and retrieving similar annotated clinical images to a given query. In the context of multi-view mammography interpretation, the concept of multi-view feature fusion has recently been studied to improve retrieval performance while effectively guaranteeing the complementarity of both MLO and CC views. However, conventional multi-view feature fusion makes descriptors long and lacks to take into consideration the relationship between descriptors. To deal with this issue, we propose two hierarchical multi-view feature fusion methods, for multi-view mammogram retrieval, based on the Canonical Correlation Analysis (CCA), which is the most commonly used multivariate parametric test. In fact, we have adapted CCA to determine the relationship between two descriptors by processing latent correlation factors. Moreover, after extracting descriptors for each view, a comparative study of texture and shape fusion descriptors is proposed in order to identify the more discriminative features for multi-view mammogram retrieval. Then, a query-dependent distance metric preserving both visual resemblance and semantic similarity is carried out to dynamically determine the more appropriate distance measure for each query image. Extensive experiments on the challenging Curated Breast Imaging Subset of Digital Database for Screening Mammography (CBIS-DDSM) have demonstrated the effectiveness of the proposed hierarchical multi-view feature fusion for mammogram retrieval, which outperforms the performance achieved either by conventional fused information or by single view information. To improve the transparency of our paper, the source code of the proposed method and the related dataset (including readme files) are publicly accessible through the following GitHub link: https://github.com/ABDERRAHIMMAR/Multi-View-Feature-Fusion-for-Mammogram-Retrieval . This open-access resource empowers researchers and practitioners to delve deeper into our methodology, fostering collaboration and advancements in the field of computer-aided diagnosis and medical image analysis.

Artificial intelligence as a second reader for screening mammography

Abstract Background Artificial intelligence (AI) has shown promise in mammography interpretation, and its use as a second reader in breast cancer screening may reduce the burden on health care systems. Purpose To evaluate the performance differences between routine double read and an AI as a second reader workflow (AISR), where the second reader is replaced with AI. Materials and Methods A cohort of patients undergoing routine breast cancer screening at a single center with mammography was retrospectively collected between 2005 and 2021. A model developed on US and UK data was fine-tuned on Japanese data. We subsequently performed a reader study with 10 qualified readers with varied experience (5 reader pairs), comparing routine double read to an AISR workflow. Results A “test set” of 4,059 women (mean age, 56 ± 14 years; 157 positive, 3,902 negative) was collected, with 278 (mean age 55 ± 13 years; 90 positive, 188 negative) evaluated for the reader study. We demonstrate an area under the curve =.84 (95% confidence interval [CI], 0.805-0.881) on the test set, with no significant difference to decisions made in clinical practice (P = .32). Compared with routine double reading, in the AISR arm, sensitivity improved by 7.6% (95% CI, 3.80-11.4; P = .00004) and specificity decreased 3.4% (1.42-5.43; P = .0016), with 71% (212/298) of scans no longer requiring input from a second reader. Variation in recall decision between reader pairs improved from a Cohen kappa of κ = .65 (96% CI, 0.61-0.68) to κ = .74 (96% CI, 0.71-0.77) in the AISR arm. Conclusion AISR improves sensitivity, reduces variability and decreases workload compared to routine dual screening.

Early Detection of Breast Cancer using Deep Learning in Mammograms

Breast cancer still poses a serious threat to world health, needing creative approaches to early identification in order to improve patient outcomes. This study investigates the potential of deep learning methods to improve the precision and effectiveness of mammography interpretation for the identification of breast cancer. In this paper proposed, a convolutional neural network (CNN) architecture, ResNet50, is created and trained on a sizable data set of annotated mammograms. The CNN is made to automatically identify and extract pertinent elements, such as microcalcifications, masses, and architectural distortions,that may be symptomatic of possible cancers. The model develops the ability to distinguish between benign and malignant instances through an iterative process of training and validation, finally displaying a high level of discriminatory accuracy. The paper findings show that the deep learning model outperforms conventional mammography interpretations in terms of sensitivity and specificity for detecting breast cancer. Furthermore, the model's potential for use in actual clinical settings is highlighted by its generalizability across a range of patient demographics and imaging technologies.This study represents a big step in improving radiologists' capacity for breast cancer early detection. Our deep learning-based architecture has promise for improving the screening procedure and potentially decreasing the difficulties brought on by radiologist shortages by lowering false positives, improving accuracy, and offering quick analysis. By utilising cutting-edge technology to enable prompt and efficient detection, this study contributes to continuing efforts by the international healthcare community to improve breast cancer outcomes.

AI analytics can be used as imaging biomarkers for predicting invasive upgrade of ductal carcinoma in situ

ObjectivesTo evaluate whether the quantitative abnormality scores provided by artificial intelligence (AI)-based computer-aided detection/diagnosis (CAD) for mammography interpretation can be used to predict invasive upgrade in ductal carcinoma in situ (DCIS) diagnosed on percutaneous biopsy.MethodsFour hundred forty DCIS in 420 women (mean age, 52.8 years) diagnosed via percutaneous biopsy from January 2015 to December 2019 were included. Mammographic characteristics were assessed based on imaging features (mammographically occult, mass/asymmetry/distortion, calcifications only, and combined mass/asymmetry/distortion with calcifications) and BI-RADS assessments. Routine pre-biopsy 4-view digital mammograms were analyzed using AI-CAD to obtain abnormality scores (AI-CAD score, ranging 0–100%). Multivariable logistic regression was performed to identify independent predictive mammographic variables after adjusting for clinicopathological variables. A subgroup analysis was performed with mammographically detected DCIS.ResultsOf the 440 DCIS, 117 (26.6%) were upgraded to invasive cancer. Three hundred forty-one (77.5%) DCIS were detected on mammography. The multivariable analysis showed that combined features (odds ratio (OR): 2.225, p = 0.033), BI-RADS 4c or 5 assessments (OR: 2.473, p = 0.023 and OR: 5.190, p < 0.001, respectively), higher AI-CAD score (OR: 1.009, p = 0.007), AI-CAD score ≥ 50% (OR: 1.960, p = 0.017), and AI-CAD score ≥ 75% (OR: 2.306, p = 0.009) were independent predictors of invasive upgrade. In mammographically detected DCIS, combined features (OR: 2.194, p = 0.035), and higher AI-CAD score (OR: 1.008, p = 0.047) were significant predictors of invasive upgrade.ConclusionThe AI-CAD score was an independent predictor of invasive upgrade for DCIS. Higher AI-CAD scores, especially in the highest quartile of ≥ 75%, can be used as an objective imaging biomarker to predict invasive upgrade in DCIS diagnosed with percutaneous biopsy.Critical relevance statementNoninvasive imaging features including the quantitative results of AI-CAD for mammography interpretation were independent predictors of invasive upgrade in lesions initially diagnosed as ductal carcinoma in situ via percutaneous biopsy and therefore may help decide the direction of surgery before treatment.Key points• Predicting ductal carcinoma in situ upgrade is important, yet there is a lack of conclusive non-invasive biomarkers.• AI-CAD scores—raw numbers, ≥ 50%, and ≥ 75%—predicted ductal carcinoma in situ upgrade independently.• Quantitative AI-CAD results may help predict ductal carcinoma in situ upgrade and guide patient management.Graphical

Interval breast cancer: radiological surveillance in screening Egyptian population

BackgroundInterval cancers might be divided into true negative interval cancer where a new lesion is detected that no sign of disease could be detected on previous screening mammogram. For false-negative interval cancers, those missed for overt symptoms and those missed for mild or undetectable ones, this includes interpretive error as benign interval cancer (benign mimics), subtle changes, masked carcinoma or slowly growing or patient factors, such as the dense breast parenchyma. Technical failure interval cancer hampered the reader to discover the abnormality. The aim of this study was to relate the risk factors for the development of the interval breast cancer such as breast density, positive family history of breast cancer in Egyptian population. Highlight the causes of missed breast cancer in order to overcome it in the future radiological interpretation.MethodsA total of 74,546 screening mammographic examinations were performed in the radiology department at specialized breast cancer center in the period between 2015 and 2021 with about 844 pathologically proved malignant cases. Out of the 844 pathologically proven breast cancer cases, 185 breast cancer patients were interval breast cancer having reported normal examination in the previous year, 88 were true interval breast cancer and 97 were missed on mammography and detected later on. The cases were subjected to full-field digital mammogram (FFDM), complementary ultrasound, contrast-enhanced digital mammography (CEDM) and magnetic resonance imaging (MRI) in some cases, and all cases were histopathologically proven by either fine needle aspiration, core needle biopsy (CNB) or open biopsy.ResultsThe mean age of the patients was 53.5 years (range 36–83 years). The overall breast cancer detection rate was 11 per 1000 women. The 185 interval cancers were present at a rate of 3.8 per 1000 women. True negative interval breast cancers where 88 cancers were not present in retrospect on screening mammograms, 17 cases present with benign findings (benign mimics mass or calcifications) and 80 cancers were missed cancers. Analysis of the cause of missed carcinoma revealed patient-related, tumor-related, or provider-related factors. Tumor factors were the most commonly encountered, accounting for 49.5% (48 cases), then provider factors in 25.8% (25 cases) while the patient factors were the least commonly encountered in 24.7% (24 cases). Recorded positive family history found in about 35% (31cases) of the true interval breast cancers.ConclusionsAlthough mammography is the standard for detecting early breast cancer, some cancers can be missed due to various causes. Mammographic interpretation must meet high standards to reduce missed cancers. Radiologists should carefully assess screening views and order additional imaging if needed. Palpable lesions and clinical data should be further examined with ultrasound and contrast imaging if necessary. Always compare current images with previous examinations to check for any changes in lesion size. When one pathology is found, search for additional lesions.

Impact of Artificial Intelligence on Mammography Interpretation by Breast Radiologists, Non-Breast Radiologists, and Senior Residents

Background: Artificial intelligence (AI) is recognized to have tremendous potential to revolutionize breast cancer management through mammography. However, the extent of its impact on radiologists with different levels of experience remains largely unexplored. Therefore, this study aimed to comprehensively show how AI could assist radiologists of varying expertise including breast and non-breast radiologists, as well as senior residents, in performing mammogram interpretation.Methods: This retrospective study analyzed eligible mammograms from Cipto Mangunkusumo Hospital between January 2017 and March 2021. Mammographic readings were conducted independently by two breast radiologists, two from other subspecialties, and three senior residents, all blinded to clinical information. AI standalone performance, as well as radiologists with and without AI assistance, was measured. Results: The results showed that a total of 886 eligible mammograms were analyzed. AI standalone performance, assessed using ROC curve analysis, yielded an AUC of 0.946 (95% CI, 0.925–0.967) with sensitivity and specificity of 90.1% and 93.6%, respectively. AI assistance significantly improved the sensitivity and specificity of all radiologists, regardless of experience level, with a median increase of 19.4% (IQR, 10.4–33.5%) and 12.1% (IQR, 5.2–16.2%), respectively. Moreover, there was a trend toward a higher increase with AI assistance in dense compared to fatty breasts.Conclusions: AI proved to be a highly effective diagnostic supplement for radiologists across varying experience levels, specifically in non-breast radiologists, offering the potential to add even greater value in cases of dense breast tissue. The results were derived from a national referral tertiary hospital that generally received many breast cancer cases referred from other hospitals for further treatment. Therefore, further studies incorporating different levels of hospitals were needed.

Impact of Interruptions During Screening Mammography on Physician Well-Being and Patient Care

ObjectiveTo assess the impact of interruptions on radiologists’ efficiency, accuracy, and job satisfaction in interpreting screening mammograms. MethodsThis institutional review board–approved retrospective reader study recruited nine breast radiologists from a single academic institution [name withheld] to interpret 150 screening mammograms performed between December 1, 2008, and December 31, 2015 under two different reading conditions, as follows: (1) uninterrupted batch reading and (2) interrupted reading. The 150 cases consisted of 125 normal mammograms and 25 mammograms with subtle breast cancers. Cases were divided into two groups of 75 cases each (cohort 1 and cohort 2), with a comparable distribution of cancer cases. Four rounds of 75 cases each were conducted with a 6-week washout period between rounds 2 and 3. After completing each interpretation session, readers completed a seven-question survey, assessing perceptions of mental and physical effort, level of frustration, and performance satisfaction. Clinical performance metrics (reading time, recall rate, sensitivity, specificity, accuracy, and positive predictive value 1) were calculated. ResultsRecall rates were significantly (P = .04) higher during interrupted reading sessions (35.4%) than they were during uninterrupted batch reading sessions (31.4%). Accuracy was significantly (P = .049) worse in the interrupted reading sessions (69.5%), compared with uninterrupted sessions (73.6%). Differences in overall image interpretation times were not statistically significant (P = .065). Compared with uninterrupted batch reading sessions, readers during interrupted sessions reported feeling busier (P < .001), encountered higher levels of cognitive demand (P = .005), experienced elevated levels of physical fatigue (P = .004), and expressed lower levels of satisfaction with their performance (P = .041). ConclusionInterruptions during interpretation of screening mammography have deleterious effects on physician performance and their sense of well-being.

Improving mammography interpretation for both novice and experienced readers: a comparative study of two commercial artificial intelligence software.

To evaluate the improvement of mammography interpretation for novice and experienced radiologists assisted by two commercial AI software. We compared the performance of two AI software (AI-1 and AI-2) in two experienced and two novice readers for 200 mammographic examinations (80 cancer cases). Two reading sessions were conducted within 4weeks. The readers rated the likelihood of malignancy (range, 1-7) and the percentage probability of malignancy (range, 0-100%), with and without AI assistance. Differences in AUROC, sensitivity, and specificity were analyzed. Mean AUROC increased in both novice (0.86 to 0.90 with AI-1 [p = 0.005]; 0.91 with AI-2 [p < 0.001]) and experienced readers (0.87 to 0.92 with AI-1 [p < 0.001]; 0.90 with AI-2 [p = 0.004]). Sensitivities increased from 81.3 to 88.8% with AI-1 (p = 0.027) and to 91.3% with AI-2 (p = 0.005) in novice readers, and from 81.9 to 90.6% with AI-1 (p = 0.001) and to 87.5% with AI-2 (p = 0.016) in experienced readers. Specificity did not decrease significantly in both novice (p > 0.999, both) and experienced readers (p > 0.999 with AI-1 and 0.282 with AI-2). There was no significant difference in the performance change depending on the type of AI software (p > 0.999). Commercial AI software improved the diagnostic performance of both novice and experienced readers. The type of AI software used did not significantly impact performance changes. Further validation with a larger number of cases and readers is needed. Commercial AI software effectively aided mammography interpretation irrespective of the experience level of human readers. • Mammography interpretation remains challenging and is subject to a wide range of interobserver variability. • In this multi-reader study, two commercial AI software improved the sensitivity of mammography interpretation by both novice and experienced readers. The type of AI software used did not significantly impact performance changes. • Commercial AI software may effectively support mammography interpretation irrespective of the experience level of human readers.

Deep learning performance for detection and classification of microcalcifications on mammography

BackgroundBreast cancer screening through mammography is crucial for early detection, yet the demand for mammography services surpasses the capacity of radiologists. Artificial intelligence (AI) can assist in evaluating microcalcifications on mammography. We developed and tested an AI model for localizing and characterizing microcalcifications.MethodsThree expert radiologists annotated a dataset of mammograms using histology-based ground truth. The dataset was partitioned for training, validation, and testing. Three neural networks (AlexNet, ResNet18, and ResNet34) were trained and evaluated using specific metrics including receiver operating characteristics area under the curve (AUC), sensitivity, and specificity. The reported metrics were computed on the test set (10% of the whole dataset).ResultsThe dataset included 1,000 patients aged 21–73 years and 1,986 mammograms (180 density A, 220 density B, 380 density C, and 220 density D), with 389 malignant and 611 benign groups of microcalcifications. AlexNet achieved the best performance with 0.98 sensitivity, 0.89 specificity of, and 0.98 AUC for microcalcifications detection and 0.85 sensitivity, 0.89 specificity, and 0.94 AUC of for microcalcifications classification. For microcalcifications detection, ResNet18 and ResNet34 achieved 0.96 and 0.97 sensitivity, 0.91 and 0.90 specificity and 0.98 and 0.98 AUC, retrospectively. For microcalcifications classification, ResNet18 and ResNet34 exhibited 0.75 and 0.84 sensitivity, 0.85 and 0.84 specificity, and 0.88 and 0.92 AUC, respectively.ConclusionsThe developed AI models accurately detect and characterize microcalcifications on mammography.Relevance statementAI-based systems have the potential to assist radiologists in interpreting microcalcifications on mammograms. The study highlights the importance of developing reliable deep learning models possibly applied to breast cancer screening.Key points• A novel AI tool was developed and tested to aid radiologists in the interpretation of mammography by accurately detecting and characterizing microcalcifications.• Three neural networks (AlexNet, ResNet18, and ResNet34) were trained, validated, and tested using an annotated dataset of 1,000 patients and 1,986 mammograms.• The AI tool demonstrated high accuracy in detecting/localizing and characterizing microcalcifications on mammography, highlighting the potential of AI-based systems to assist radiologists in the interpretation of mammograms.Graphical

Reproducibility and Explainability of Deep Learning in Mammography: A Systematic Review of Literature.

Background Although abundant literature is currently available on the use of deep learning for breast cancer detection in mammography, the quality of such literature is widely variable. Purpose To evaluate published literature on breast cancer detection in mammography for reproducibility and to ascertain best practices for model design. Methods The PubMed and Scopus databases were searched to identify records that described the use of deep learning to detect lesions or classify images into cancer or noncancer. A modification of Quality Assessment of Diagnostic Accuracy Studies (mQUADAS-2) tool was developed for this review and was applied to the included studies. Results of reported studies (area under curve [AUC] of receiver operator curve [ROC] curve, sensitivity, specificity) were recorded. Results A total of 12,123 records were screened, of which 107 fit the inclusion criteria. Training and test datasets, key idea behind model architecture, and results were recorded for these studies. Based on mQUADAS-2 assessment, 103 studies had high risk of bias due to nonrepresentative patient selection. Four studies were of adequate quality, of which three trained their own model, and one used a commercial network. Ensemble models were used in two of these. Common strategies used for model training included patch classifiers, image classification networks (ResNet in 67%), and object detection networks (RetinaNet in 67%). The highest reported AUC was 0.927 ± 0.008 on a screening dataset, while it reached 0.945 (0.919-0.968) on an enriched subset. Higher values of AUC (0.955) and specificity (98.5%) were reached when combined radiologist and Artificial Intelligence readings were used than either of them alone. None of the studies provided explainability beyond localization accuracy. None of the studies have studied interaction between AI and radiologist in a real world setting. Conclusion While deep learning holds much promise in mammography interpretation, evaluation in a reproducible clinical setting and explainable networks are the need of the hour.

Analysis of Specimen Mammography with Artificial Intelligence to Predict Margin Status.

Intraoperative specimen mammography is a valuable tool in breast cancer surgery, providing immediate assessment of margins for a resected tumor. However, the accuracy of specimen mammography in detecting microscopic margin positivity is low. We sought to develop an artificial intelligence model to predict the pathologic margin status of resected breast tumors using specimen mammography. A dataset of specimen mammography images matched with pathologic margin status was collected from our institution from 2017 to 2020. The dataset was randomly split into training, validation, and test sets. Specimen mammography models pretrained on radiologic images were developed and compared with models pretrained on nonmedical images. Model performance was assessed using sensitivity, specificity, and area under the receiver operating characteristic curve (AUROC). The dataset included 821 images, and 53% had positive margins. For three out of four model architectures tested, models pretrained on radiologic images outperformed nonmedical models. The highest performing model, InceptionV3, showed sensitivity of 84%, specificity of 42%, and AUROC of 0.71. Model performance was better among patients with invasive cancers, less dense breasts, and non-white race. This study developed and internally validated artificial intelligence models that predict pathologic margins status for partial mastectomy from specimen mammograms. The models' accuracy compares favorably with published literature on surgeon and radiologist interpretation of specimen mammography. With further development, these models could more precisely guide the extent of resection, potentially improving cosmesis and reducing reoperations.

Double-reading mammograms using artificial intelligence technologies: A new model of mass preventive examination organization

BACKGROUND: In recent years, the availability of medical datasets and technologies for software development based on artificial intelligence technology has resulted in a growth in the number of solutions for medical diagnostics, particularly mammography. Registered as a medical device, this program can interpret digital mammography, significantly saving time, material, and human resources in healthcare while ensuring the quality of mammary gland preventive studies.&#x0D; AIM: This study aims to justify the possibility and effectiveness of artificial intelligence-based software for the first interpretation of digital mammograms while maintaining the practice of a radiologists second description of X-ray images.&#x0D; MATERIALS AND METHODS: A dataset of 100 digital mammography studies (50 absence of target pathology and 50 ― presence of target pathology, with signs of malignant neoplasms) was processed by software based on artificial intelligence technology that was registered as a medical device in the Russian Federation. Receiver operating characteristic analysis was performed. Limitations of the study include the values of diagnostic accuracy metrics obtained for software based on artificial intelligence technology versions, relevant at the end of 2022.&#x0D; RESULTS: When set to 80.0% sensitivity, artificial intelligence specificity was 90.0% (95% CI, 81.798.3), and accuracy was 85.0% (95% CI, 78.092.0). When set to 100% specificity, artificial intelligence demonstrated 56.0% sensitivity (95% CI, 42.269.8) and 78.0% accuracy (95% CI, 69.986.1). When the sensitivity was set to 100%, the artificial intelligence specificity was 54.0% (95% CI, 40.267.8), and the accuracy was 77.0% (95% CI, 68.885.2). Two approaches have been proposed, providing an autonomous first interpretation of digital mammography using artificial intelligence. The first approach is to evaluate the X-ray image using artificial intelligence with a higher sensitivity than that of the double-reading mammogram by radiologists, with a comparable level of specificity. The second approach implies that artificial intelligence-based software will determine the mammogram category (absence of target pathology or presence of target pathology), indicating the degree of confidence in the obtained result, depending on the corridor into which the predicted value falls.&#x0D; CONCLUSIONS: Both proposed approaches for using artificial intelligence-based software for the autonomous first interpretation of digital mammograms can provide diagnostic quality comparable to, if not superior to, double-image reading by radiologists. The economic benefit from the practical implementation of this approach nationwide can range from 0.6 to 5.5 billion rubles annually.

Artificial intelligence-based computer-assisted detection/diagnosis (AI-CAD) for screening mammography: Outcomes of AI-CAD in the mammographic interpretation workflow

PurposeTo evaluate the stand-alone diagnostic performances of AI-CAD and outcomes of AI-CAD detected abnormalities when applied to the mammographic interpretation workflow. MethodsFrom January 2016 to December 2017, 6499 screening mammograms of 5228 women were collected from a single screening facility. Historic reads of three radiologists were used as radiologist interpretation. A commercially-available AI-CAD was used for analysis. One radiologist not involved in interpretation had retrospectively reviewed the abnormality features and assessed the significance (negligible vs. need recall) of the AI-CAD marks. Ground truth in terms of cancer, benign or absence of abnormality was confirmed according to histopathologic diagnosis or negative results on the next-round screen. ResultsOf the 6499 mammograms, 6282 (96.7%) were in the negative, 189 (2.9%) were in the benign, and 28 (0.4%) were in the cancer group. AI-CAD detected 5 (17.9%, 5 of 28) of the 9 cancers that were intially interpreted as negative. Of the 648 AI-CAD recalls, 89.0% (577 of 648) were marks seen on examinations in the negative group, and 267 (41.2%) of the AI-CAD marks were considered to be negligible. Stand-alone AI-CAD has significantly higher recall rates (10.0% vs. 3.4%, P < 0.001) with comparable sensitivity and cancer detection rates (P = 0.086 and 0.102, respectively) when compared to the radiologists’ interpretation. ConclusionAI-CAD detected 17.9% additional cancers on screening mammography that were initially overlooked by the radiologists. In spite of the additional cancer detection, AI-CAD had significantly higher recall rates in the clinical workflow, in which 89.0% of AI-CAD marks are on negative mammograms.

Using Radiomics-Based Machine Learning to Create Targeted Test Sets to Improve Specific Mammography Reader Cohort Performance: A Feasibility Study.

Mammography interpretation is challenging with high error rates. This study aims to reduce the errors in mammography reading by mapping diagnostic errors against global mammographic characteristics using a radiomics-based machine learning approach. A total of 36 radiologists from cohort A (n = 20) and cohort B (n = 16) read 60 high-density mammographic cases. Radiomic features were extracted from three regions of interest (ROIs), and random forest models were trained to predict diagnostic errors for each cohort. Performance was evaluated using sensitivity, specificity, accuracy, and AUC. The impact of ROI placement and normalization on prediction was investigated. Our approach successfully predicted both the false positive and false negative errors of both cohorts but did not consistently predict location errors. The errors produced by radiologists from cohort B were less predictable compared to those in cohort A. The performance of the models did not show significant improvement after feature normalization, despite the mammograms being produced by different vendors. Our novel radiomics-based machine learning pipeline focusing on global radiomic features could predict false positive and false negative errors. The proposed method can be used to develop group-tailored mammographic educational strategies to help improve future mammography reader performance.

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.

Effect of artificial intelligence-based computer-aided diagnosis on the screening outcomes of digital mammography: a matched cohort study.

To investigate whether artificial intelligence-based computer-aided diagnosis (AI-CAD) can improve radiologists' performance when used to support radiologists' interpretation of digital mammography (DM) in breast cancer screening. A retrospective database search identified 3158 asymptomatic Korean women who consecutively underwent screening DM between January and December 2019 without AI-CAD support, and screening DM between February and July 2020 with image interpretation aided by AI-CAD in a tertiary referral hospital using single reading. Propensity score matching was used to match the DM with AI-CAD group in a 1:1 ratio with the DM without AI-CAD group according to age, breast density, experience level of the interpreting radiologist, and screening round. Performance measures were compared with the McNemar test and generalized estimating equations. A total of 1579 women who underwent DM with AI-CAD were matched with 1579 women who underwent DM without AI-CAD. Radiologists showed higher specificity (96% [1500 of 1563] vs 91.6% [1430 of 1561]; p < 0.001) and lower abnormal interpretation rates (AIR) (4.9% [77 of 1579] vs 9.2% [145 of 1579]; p < 0.001) with AI-CAD than without. There was no significant difference in the cancer detection rate (CDR) (AI-CAD vs no AI-CAD, 8.9 vs 8.9 per 1000 examinations; p = 0.999), sensitivity (87.5% vs 77.8%; p = 0.999), and positive predictive value for biopsy (PPV3) (35.0% vs 35.0%; p = 0.999) according to AI-CAD support. AI-CAD increases the specificity for radiologists without decreasing sensitivity as a supportive tool in the single reading of DM for breast cancer screening. This study shows that AI-CAD could improve the specificity of radiologists' DM interpretation in the single reading system without decreasing sensitivity, suggesting that it can benefit patients by reducing false positive and recall rates. • In this retrospective-matched cohort study (DM without AI-CAD vs DM with AI-CAD), radiologists showed higher specificity and lower AIR when AI-CAD was used to support decision-making in DM screening. • CDR, sensitivity, and PPV for biopsy did not differ with and without AI-CAD support.