Deep Learning for Automated Triaging of 4581 Breast MRI Examinations from the DENSE Trial

Pilot study: A simple CAD-based tool to detect breast cancer on MRI of the breast

PURPOSE

This pilot-study aims to assess, whether quantitatively assessed enhancing breast tissue as a percentage of the entire breast volume can serve as an indicator of breast cancer at breast MRI and whether the contrast-agent employed affects diagnostic efficacy.

MATERIALS

This retrospective IRB-approved study, included 39 consecutive patients, that underwent two subsequent breast MRI exams for suspicious findings at conventional imaging with 0.1 mmol/kg gadobenic and gadoteric acid. Two independent readers, blinded to the histopathological outcome, assessed unenhanced and early post-contrast images using computer-assisted software (Brevis, Siemens Healthcare). Diagnostic performance was statistically determined for percentage of ipsilateral voxel volume enhancement and for percentage of contralateral enhancing voxel volume subtracted from ipsilateral enhancing voxel volume after crosstabulation with the dichotomized histological outcome (benign/malignant).

RESULTS

Ipsilateral enhancing voxel volume versus histopathological outcome resulted in an AUC of 0.707 and 0.687 for gadobenic acid, reader 1 and 2, respectively and in an AUC of 0.778 and 0.773 for gadoteric acid, reader 1 and 2, respectively. Accounting for background parenchymal enhancement by subtracting contralateral enhancing volume from ipsilateral enhancing voxel volume versus histolopathological outcome resulted in an AUC of 0.793 and 0.843 for gadobenic acid, reader 1 and 2, respectively and in an AUC of 0.692 and 0.662 for gadoteric acid, reader 1 and 2, respectively. Pairwise testing yielded no statistically significant difference both between readers and between contrast agents employed (p > 0.05).

CONCLUSION

Our proposed CAD algorithm, which quantitatively assesses enhancing breast tissue as a percentage of the entire breast volume, allows indicating the presence of breast cancer.

Familiarity, confidence and preference of artificial intelligence feedback and prompts by Australian breast cancer screening readers

Objectives This study explored the familiarity, perceptions and confidence of Australian radiology clinicians involved in reading screening mammograms, regarding artificial intelligence (AI) applications in breast cancer detection. Methods Sixty-five radiologists, breast physicians and radiology trainees participated in an online survey that consisted of 23 multiple choice questions asking about their experience and familiarity with AI products. Furthermore, the survey asked about their confidence in using AI outputs and their preference for AI modes applied in a breast screening context. Participants' responses to questions were compared using Pearson's χ 2 test. Bonferroni-adjusted significance tests were used for pairwise comparisons. Results Fifty-five percent of respondents had experience with AI in their workplaces, with automatic density measurement powered by machine learning being the most familiar AI product (69.4%). The top AI outputs with the highest ranks of perceived confidence were 'Displaying suspicious areas on mammograms with the percentage of cancer possibility' (67.8%) and 'Automatic mammogram classification (normal, benign, cancer, uncertain)' (64.6%). Radiology and breast physicians preferred using AI as second-reader mode (75.4% saying 'somewhat happy' to 'extremely happy') over triage (47.7%), pre-screening and first-reader modes (both with 26.2%) (P  < 0.001). Conclusion The majority of screen readers expressed increased confidence in utilising AI for highlighting suspicious areas on mammograms and for automatically classifying mammograms. They considered AI as an optimal second-reader mode being the most ideal use in a screening program. The findings provide valuable insights into the familiarities and expectations of radiologists and breast clinicians for the AI products that can enhance the effectiveness of the breast cancer screening programs, benefitting both healthcare professionals and patients alike.

Implementing supplementary breast cancer screening in women with dense breasts: Insights from European radiographers and radiologists

INTRODUCTION

In response to the critical need for enhancing breast cancer screening for women with dense breasts, this study explored the understanding of challenges and requirements for implementing supplementary breast cancer screening for such women among clinical radiographers and radiologists in Europe.

METHOD

Fourteen (14) semi-structured online interviews were conducted with European clinical radiologists (n = 5) and radiographers (n = 9) specializing in breast cancer screening from 8 different countries: Denmark, Finland, Greece, Italy, Malta, the Netherlands, Switzerland, United Kingdom. The interview schedule comprised questions regarding professional background and demographics and 13 key questions divided into six subgroups, namely Supplementary Imaging, Training, Resources and Guidelines, Challenges, Implementing supplementary screening and Women's Perspective. Data analysis followed the six phases of reflexive thematic analysis.

RESULTS

Six significant themes emerged from the data analysis: Understanding and experiences of supplementary imaging for women with dense breasts; Challenges and requirements related to training among clinical radiographers and radiologists; Awareness among radiographers and radiologists of guidelines on imaging women with dense breasts; Challenges to implement supplementary screening; Predictors of Implementing Supplementary screening; Views of radiologists and radiographers on women's perception towards supplementary screening.

CONCLUSION

The interviews with radiographers and radiologists provided valuable insights into the challenges and potential strategies for implementing supplementary breast cancer screening. These challenges included patient and staff related challenges. Implementing multifaceted solutions such as Artificial Intelligence integration, specialized training and resource investment can address these challenges and promote the successful implementation of supplementary screening. Further research and collaboration are needed to refine and implement these strategies effectively.

IMPLICATIONS FOR PRACTICE

This study highlights the urgent need for specialized training programs and dedicated resources to enhance supplementary breast cancer screening for women with dense breasts in Europe. These resources include advanced imaging technologies, such as MRI or ultrasound, and specialized software for image analysis. Moreover, further research is imperative to refine screening protocols and evaluate their efficacy and cost-effectiveness, based on the findings of this study.

AI Applications to Breast MRI: Today and Tomorrow

In breast imaging, there is an unrelenting increase in the demand for breast imaging services, partly explained by continuous expanding imaging indications in breast diagnosis and treatment. As the human workforce providing these services is not growing at the same rate, the implementation of artificial intelligence (AI) in breast imaging has gained significant momentum to maximize workflow efficiency and increase productivity while concurrently improving diagnostic accuracy and patient outcomes. Thus far, the implementation of AI in breast imaging is at the most advanced stage with mammography and digital breast tomosynthesis techniques, followed by ultrasound, whereas the implementation of AI in breast magnetic resonance imaging (MRI) is not moving along as rapidly due to the complexity of MRI examinations and fewer available dataset. Nevertheless, there is persisting interest in AI-enhanced breast MRI applications, even as the use of and indications of breast MRI continue to expand. This review presents an overview of the basic concepts of AI imaging analysis and subsequently reviews the use cases for AI-enhanced MRI interpretation, that is, breast MRI triaging and lesion detection, lesion classification, prediction of treatment response, risk assessment, and image quality. Finally, it provides an outlook on the barriers and facilitators for the adoption of AI in breast MRI. LEVEL OF EVIDENCE: 5 TECHNICAL EFFICACY: Stage 6.

Artificial intelligence in breast imaging: potentials and challenges

Breast cancer, which is the most common type of malignant tumor among humans, is a leading cause of death in females. Standard treatment strategies, including neoadjuvant chemotherapy, surgery, postoperative chemotherapy, targeted therapy, endocrine therapy, and radiotherapy, are tailored for individual patients. Such personalized therapies have tremendously reduced the threat of breast cancer in females. Furthermore, early imaging screening plays an important role in reducing the treatment cycle and improving breast cancer prognosis. The recent innovative revolution in artificial intelligence (AI) has aided radiologists in the early and accurate diagnosis of breast cancer. In this review, we introduce the necessity of incorporating AI into breast imaging and the applications of AI in mammography, ultrasonography, magnetic resonance imaging, and positron emission tomography/computed tomography based on published articles since 1994. Moreover, the challenges of AI in breast imaging are discussed.

Deep learning-based recurrence detector on magnetic resonance scans in nasopharyngeal carcinoma: A multicenter study

OBJECTIVES

Accuracy in the detection of recurrent nasopharyngeal carcinoma (NPC) on follow-up magnetic resonance (MR) scans needs to be improved.

MATERIAL AND METHODS

A total of 5 035 follow-up MR scans from 5 035 survivors with treated NPC between April 2007 and July 2020 were retrospectively collected from three cancer centers for developing and evaluating the deep learning (DL) model MODERN (MR-based Deep learning model for dEtecting Recurrent Nasopharyngeal carcinoma). In a reader study with 220 scans, the accuracy of two radiologists in detecting recurrence on scans with vs without MODERN was evaluated. The performance was measured using the area under the receiver operating characteristic curve (ROC-AUC) and accuracy with a 95% confidence interval (CI).

RESULTS

MODERN exhibited sound performance in the validation cohort (internal: ROC-AUC, 0.88, 95% CI, 0.86-0.90; external 1: ROC-AUC, 0.88, 95% CI, 0.86-0.90; external 2: ROC-AUC, 0.85, 95% CI, 0.82-0.88). In a reader study, MODERN alone achieved reliable accuracy compared to that of radiologists (MODERN: 84.1%, 95% CI, 79.3%-88.9%; competent: 80.9%, 95% CI, 75.7%-86.1%, P < 0.001; expert: 85.9%, 95% CI, 81.3%-90.5%, P < 0.001). The accuracy of radiologists was boosted by the MODERN score (competent with MODERN score: 84.6%, 95% CI, 79.8%-89.3%, P < 0.001; expert with MODERN score: 87.7%, 95% CI, 83.4%-92.1%, P < 0.001).

CONCLUSION

We developed a DL model for recurrence detection with reliable performance. Computer-human collaboration has the potential to refine the workflow in interpreting surveillant MR scans among patients with treated NPC.

DI-UNet: dual-branch interactive U-Net for skin cancer image segmentation

PURPOSE

Skin disease is a prevalent type of physical ailment that can manifest in multitude of forms. Many internal diseases can be directly reflected on the skin, and if left unattended, skin diseases can potentially develop into skin cancer. Accurate and effective segmentation of skin lesions, especially melanoma, is critical for early detection and diagnosis of skin cancer. However, the complex color variations, boundary ambiguity, and scale variations in skin lesion regions present significant challenges for precise segmentation.

METHODS

We propose a novel approach for melanoma segmentation using a dual-branch interactive U-Net architecture. Two distinct sampling strategies are simultaneously integrated into the network, creating a vertical dual-branch structure. Meanwhile, we introduce a novel dual-channel symmetrical convolution block (DCS-Conv), which employs a symmetrical design, enabling the network to exhibit a horizontal dual-branch structure. The combination of the vertical and horizontal distribution of the dual-branch structure enhances both the depth and width of the network, providing greater diversity and rich multiscale cascade features. Additionally, this paper introduces a novel module called the residual fuse-and-select module (RFS module), which leverages self-attention mechanisms to focus on the specific skin cancer features and reduce irrelevant artifacts, further improving the segmentation accuracy.

RESULTS

We evaluated our approach on two publicly skin cancer datasets, ISIC2016 and PH2, and achieved state-of-the-art results, surpassing previous outcomes in terms of segmentation accuracy and overall performance.

CONCLUSION

Our proposed approach holds tremendous potential to aid dermatologists in clinical decision-making.

[Echo Train Length (ETL) of Fluid-attenuated Inversion Recovery (FLAIR) and Extraction Volume of White Matter Hyperintensity Volume in Automated White Matter Signal Analysis]

PURPOSE

To investigate whether the volume of white matter hyperintensity (WMH) extracted from FLAIR images changes when the imaging parameters of the original images are changed.

METHODS

Seven healthy volunteers were imaged by changing the imaging parameter ETL of FLAIR images, and WMHs were extracted and their volumes were calculated by the automatic extraction software. The results were statistically analyzed to examine the relationship (Experiment 1). Simulated images with different SNRs were created by adding white noise to four examples of healthy volunteer images. The SNR of the simulated images simulated the SNR of the measured images of different ETLs. The WMH was extracted from the simulated images and its volume was calculated using the automatic extraction software (Experiment 2).

RESULTS

Experiment 1 showed that there was no significant difference between FLAIR imaging parameters and WMH volume in automatic white matter signal analysis, except for some conditions. Experiment 2 showed that as the SNR of the original image decreased, the volume of high white matter signal extracted decreased.

CONCLUSION

In automatic white matter signal analysis, WMH was shown to be small when the ETL of the FLAIR sequence was larger than normal and/or the SNR of the image was low.

Criticality and clinical department prediction of ED patients using machine learning based on heterogeneous medical data

PROBLEM

Emergency triage faces multiple challenges, including limited medical resources and inadequate manual triage nurses, which cause incorrect triage, overcrowding in the emergency department (ED), and long patient waiting time.

OBJECTIVE

This paper aims to propose and validate an accurate and efficient artificial intelligence-based method for effectively ED triage and alleviating the pressure on medical resources.

METHODS

We propose two novel machine learning models, TransNet and TextRNN, for predicting patient severity levels and clinical departments using heterogeneous medical data in ED triage. Our models employ a parallel structure for feature extraction and incorporate an attention mechanism to extract essential information from the fused features, enabling accurate predictions. The models analyze the triage data (2020-2022) from the ED of Beijing University People's Hospital, incorporating variables (demographics, triage vital signs, and chief complaints) to identify patient severity levels and clinical departments. We performed data cleaning, categorization, and encoding first. Then, we divided the available data into a training set (56%), a validation set (24%), and a test set (20%) by random sampling. Finally, our models underwent 5-fold cross-validation and were compared with other state-of-the-art models.

RESULTS

We comprehensively evaluated the proposed models against various Recurrent Neural Networks (RNN), Convolutional Neural Networks (CNN), Traditional Machine Learning (TML), and Transformer-based (TF) models, achieving excellent performance in predicting triage outcomes. Specifically, TextRNN achieved a prediction success rate of 86.23% [85.86-86.70] for severity levels and 94.30% [94.00-94.46] for clinical departments among 161,198 ED visits. Moreover, TransNet demonstrated higher sensitivities of 84.08% and 90.05% for severity levels and clinical departments, respectively, with specificities of 76.48% and 95.16%. The accuracy of our model is 0.87%, 0.18%, 4.29%, and 1.96%, higher than that of the above four family models on average. Furthermore, our method significantly reduced under-triage by 12.06% and over-triage by 17.92% compared to manual triage.

CONCLUSIONS

Experimental results demonstrated that the proposed models fuse heterogeneous medical data in the triage process, successfully predicting patients' triage outcomes. Our models can improve triage efficiency, reduce the under/over-triage rate, and provide physicians with valuable decision-making support.

Automated Triage of Screening Breast MRI Examinations in High-Risk Women Using an Ensemble Deep Learning Model

OBJECTIVES

The aim of the study is to develop and evaluate the performance of a deep learning (DL) model to triage breast magnetic resonance imaging (MRI) findings in high-risk patients without missing any cancers.

MATERIALS AND METHODS

In this retrospective study, 16,535 consecutive contrast-enhanced MRIs performed in 8354 women from January 2013 to January 2019 were collected. From 3 New York imaging sites, 14,768 MRIs were used for the training and validation data set, and 80 randomly selected MRIs were used for a reader study test data set. From 3 New Jersey imaging sites, 1687 MRIs (1441 screening MRIs and 246 MRIs performed in recently diagnosed breast cancer patients) were used for an external validation data set. The DL model was trained to classify maximum intensity projection images as "extremely low suspicion" or "possibly suspicious." Deep learning model evaluation (workload reduction, sensitivity, specificity) was performed on the external validation data set, using a histopathology reference standard. A reader study was performed to compare DL model performance to fellowship-trained breast imaging radiologists.

RESULTS

In the external validation data set, the DL model triaged 159/1441 of screening MRIs as "extremely low suspicion" without missing a single cancer, yielding a workload reduction of 11%, a specificity of 11.5%, and a sensitivity of 100%. The model correctly triaged 246/246 (100% sensitivity) of MRIs in recently diagnosed patients as "possibly suspicious." In the reader study, 2 readers classified MRIs with a specificity of 93.62% and 91.49%, respectively, and missed 0 and 1 cancer, respectively. On the other hand, the DL model classified MRIs with a specificity of 19.15% and missed 0 cancers, highlighting its potential use not as an independent reader but as a triage tool.

CONCLUSIONS

Our automated DL model triages a subset of screening breast MRIs as "extremely low suspicion" without misclassifying any cancer cases. This tool may be used to reduce workload in standalone mode, to shunt low suspicion cases to designated radiologists or to the end of the workday, or to serve as base model for other downstream AI tools.

Deep learning for risk-based stratification of cognitively impaired individuals

Quantifying the risk of progression to Alzheimer's disease (AD) could help identify persons who could benefit from early interventions. We used data from the Alzheimer's Disease Neuroimaging Initiative (ADNI, n = 544, discovery cohort) and the National Alzheimer's Coordinating Center (NACC, n = 508, validation cohort), subdividing individuals with mild cognitive impairment (MCI) into risk groups based on cerebrospinal fluid amyloid-β levels and identifying differential gray matter patterns. We then created models that fused neural networks with survival analysis, trained using non-parcellated T1-weighted brain MRIs from ADNI data, to predict the trajectories of MCI to AD conversion within the NACC cohort (integrated Brier score: 0.192 [discovery], and 0.108 [validation]). Using modern interpretability techniques, we verified that regions important for model prediction are classically associated with AD. We confirmed AD diagnosis labels using postmortem data. We conclude that our framework provides a strategy for risk-based stratification of individuals with MCI and for identifying regions key for disease prognosis.

Patient Graph Deep Learning to Predict Breast Cancer Molecular Subtype

Breast cancer is a heterogeneous disease consisting of a diverse set of genomic mutations and clinical characteristics. The molecular subtypes of breast cancer are closely tied to prognosis and therapeutic treatment options. We investigate using deep graph learning on a collection of patient factors from multiple diagnostic disciplines to better represent breast cancer patient information and predict molecular subtype. Our method models breast cancer patient data into a multi-relational directed graph with extracted feature embeddings to directly represent patient information and diagnostic test results. We develop a radiographic image feature extraction pipeline to produce vector representation of breast cancer tumors in DCE-MRI and an autoencoder-based genomic variant embedding method to map variant assay results to a low-dimensional latent space. We leverage related-domain transfer learning to train and evaluate a Relational Graph Convolutional Network to predict the probabilities of molecular subtypes for individual breast cancer patient graphs. Our work found that utilizing information from multiple multimodal diagnostic disciplines improved the model's prediction results and produced more distinct learned feature representations for breast cancer patients. This research demonstrates the capabilities of graph neural networks and deep learning feature representation to perform multimodal data fusion and representation in the breast cancer domain.

Assessing Quantitative Parenchymal Features at Baseline Dynamic Contrast-enhanced MRI and Cancer Occurrence in Women with Extremely Dense Breasts

Background Automated identification of quantitative breast parenchymal enhancement features on dynamic contrast-enhanced (DCE) MRI scans could provide added value in assessment of breast cancer risk in women with extremely dense breasts. Purpose To automatically identify quantitative properties of the breast parenchyma on baseline DCE MRI scans and assess their association with breast cancer occurrence in women with extremely dense breasts. Materials and Methods This study represents a secondary analysis of the Dense Tissue and Early Breast Neoplasm Screening trial. MRI was performed in eight hospitals between December 2011 and January 2016. After segmentation of fibroglandular tissue, quantitative features (including volumetric density, volumetric morphology, and enhancement characteristics) of the parenchyma were extracted from baseline MRI scans. Principal component analysis was used to identify parenchymal measures with the greatest variance. Multivariable Cox proportional hazards regression was applied to assess the association between breast cancer occurrence and quantitative parenchymal features, followed by stratification of significant features into tertiles. Results A total of 4553 women (mean age, 55.7 years ± 6 [SD]) with extremely dense breasts were included; of these women, 122 (3%) were diagnosed with breast cancer. Five principal components representing 96% of the variance were identified, and the component explaining the greatest independent variance (42%) consisted of MRI features relating to volume of enhancing parenchyma. Multivariable analysis showed that volume of enhancing parenchyma was associated with breast cancer occurrence (hazard ratio [HR], 1.09; 95% CI: 1.01, 1.18; P = .02). Additionally, women in the high tertile of volume of enhancing parenchyma showed a breast cancer occurrence twice that of women in the low tertile (HR, 2.09; 95% CI: 1.25, 3.61; P = .005). Conclusion In women with extremely dense breasts, a high volume of enhancing parenchyma on baseline DCE MRI scans was associated with increased occurrence of breast cancer as compared with a low volume of enhancing parenchyma. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Grimm in this issue.

Global development of artificial intelligence in cancer field: a bibliometric analysis range from 1983 to 2022

Background

Artificial intelligence (AI) is widely applied in cancer field nowadays. The aim of this study is to explore the hotspots and trends of AI in cancer research.

Methods

The retrieval term includes four topic words ("tumor," "cancer," "carcinoma," and "artificial intelligence"), which were searched in the database of Web of Science from January 1983 to December 2022. Then, we documented and processed all data, including the country, continent, Journal Impact Factor, and so on using the bibliometric software.

Results

A total of 6,920 papers were collected and analyzed. We presented the annual publications and citations, most productive countries/regions, most influential scholars, the collaborations of journals and institutions, and research focus and hotspots in AI-based cancer research.

Conclusion

This study systematically summarizes the current research overview of AI in cancer research so as to lay the foundation for future research.

'Earlier than Early' Detection of Breast Cancer in Israeli BRCA Mutation Carriers Applying AI-Based Analysis to Consecutive MRI Scans

Female BRCA1/BRCA2 (=BRCA) pathogenic variants (PVs) carriers are at a substantially higher risk for developing breast cancer (BC) compared with the average risk population. Detection of BC at an early stage significantly improves prognosis. To facilitate early BC detection, a surveillance scheme is offered to BRCA PV carriers from age 25-30 years that includes annual MRI based breast imaging. Indeed, adherence to the recommended scheme has been shown to be associated with earlier disease stages at BC diagnosis, more in-situ pathology, smaller tumors, and less axillary involvement. While MRI is the most sensitive modality for BC detection in BRCA PV carriers, there are a significant number of overlooked or misinterpreted radiological lesions (mostly enhancing foci), leading to a delayed BC diagnosis at a more advanced stage. In this study we developed an artificial intelligence (AI)-network, aimed at a more accurate classification of enhancing foci, in MRIs of BRCA PV carriers, thus reducing false-negative interpretations. Retrospectively identified foci in prior MRIs that were either diagnosed as BC or benign/normal in a subsequent MRI were manually segmented and served as input for a convolutional network architecture. The model was successful in classification of 65% of the cancerous foci, most of them triple-negative BC. If validated, applying this scheme routinely may facilitate 'earlier than early' BC diagnosis in BRCA PV carriers.

A YOLO-based AI system for classifying calcifications on spot magnification mammograms

OBJECTIVES

Use of an AI system based on deep learning to investigate whether the system can aid in distinguishing malignant from benign calcifications on spot magnification mammograms, thus potentially reducing unnecessary biopsies.

METHODS

In this retrospective study, we included public and in-house datasets with annotations for the calcifications on both craniocaudal and mediolateral oblique vies, or both craniocaudal and mediolateral views of each case of mammograms. All the lesions had pathological results for correlation. Our system comprised an algorithm based on You Only Look Once (YOLO) named adaptive multiscale decision fusion module. The algorithm was pre-trained on a public dataset, Curated Breast Imaging Subset of Digital Database for Screening Mammography (CBIS-DDSM), then re-trained and tested on the in-house dataset of spot magnification mammograms. The performance of the system was investigated by receiver operating characteristic (ROC) analysis.

RESULTS

We included 1872 images from 753 calcification cases (414 benign and 339 malignant) from CBIS-DDSM. From the in-house dataset, 636 cases (432 benign and 204 malignant) with 1269 spot magnification mammograms were included, with all lesions being recommended for biopsy by radiologists. The area under the ROC curve for our system on the in-house testing dataset was 0.888 (95% CI 0.868-0.908), with a sensitivity of 88.4% (95% CI 86.9-8.99%), specificity of 80.8% (95% CI 77.6-84%), and an accuracy of 84.6% (95% CI 81.8-87.4%) at the optimal cutoff value. Using the system with two views of spot magnification mammograms, 80.8% benign biopsies could be avoided.

CONCLUSION

The AI system showed good accuracy for classification of calcifications on spot magnification mammograms which were all categorized as suspicious by radiologists, thereby potentially reducing unnecessary biopsies.

Validation of Combined Deep Learning Triaging and Computer-Aided Diagnosis in 2901 Breast MRI Examinations From the Second Screening Round of the Dense Tissue and Early Breast Neoplasm Screening Trial

OBJECTIVES

Computer-aided triaging (CAT) and computer-aided diagnosis (CAD) of screening breast magnetic resonance imaging have shown potential to reduce the workload of radiologists in the context of dismissing normal breast scans and dismissing benign disease in women with extremely dense breasts. The aim of this study was to validate the potential of integrating CAT and CAD to reduce workload and workup on benign lesions in the second screening round of the DENSE trial, without missing cancer.

METHODS

We included 2901 breast magnetic resonance imaging scans, obtained from 8 hospitals in the Netherlands. Computer-aided triaging and CAD were previously developed on data from the first screening round. Computer-aided triaging dismissed examinations without lesions. Magnetic resonance imaging examinations triaged to radiological reading were counted and subsequently processed by CAD. The number of benign lesions correctly classified by CAD was recorded. The false-positive fraction of the CAD was compared with that of unassisted radiological reading in the second screening round. Receiver operating characteristics (ROC) analysis was performed and the generalizability of CAT and CAD was assessed by comparing results from first and second screening rounds.

RESULTS

Computer-aided triaging dismissed 950 of 2901 (32.7%) examinations with 49 lesions in total; none were malignant. Subsequent CAD classified 132 of 285 (46.3%) lesions as benign without misclassifying any malignant lesion. Together, CAT and CAD yielded significantly fewer false-positive lesions, 53 of 109 (48.6%) and 89 of 109 (78.9%), respectively ( P = 0.001), than radiological reading alone. Computer-aided triaging had a smaller area under the ROC curve in the second screening round compared with the first, 0.83 versus 0.76 ( P = 0.001), but this did not affect the negative predictive value at the 100% sensitivity operating threshold. Computer-aided diagnosis was not associated with significant differences in area under the ROC curve (0.857 vs 0.753, P = 0.08). At the operating thresholds, the specificities of CAT (39.7% vs 41.0%, P = 0.70) and CAD (41.0% vs 38.2%, P = 0.62) were successfully reproduced in the second round.

CONCLUSION

The combined application of CAT and CAD showed potential to reduce workload of radiologists and to reduce number of biopsies on benign lesions. Computer-aided triaging (CAT) correctly dismissed 950 of 2901 (32.7%) examinations with 49 lesions in total; none were malignant. Subsequent computer-aided diagnosis (CAD) classified 132 of 285 (46.3%) lesions as benign without misclassifying any malignant lesion. Together, CAT and CAD yielded significantly fewer false-positive lesions, 53 of 109 (48.6%) and 89 of 109 (78.9%), respectively ( P = 0.001), than radiological reading alone.

Applying Deep Learning for Breast Cancer Detection in Radiology

Recent advances in deep learning have enhanced medical imaging research. Breast cancer is the most prevalent cancer among women, and many applications have been developed to improve its early detection. The purpose of this review is to examine how various deep learning methods can be applied to breast cancer screening workflows. We summarize deep learning methods, data availability and different screening methods for breast cancer including mammography, thermography, ultrasound and magnetic resonance imaging. In this review, we will explore deep learning in diagnostic breast imaging and describe the literature review. As a conclusion, we discuss some of the limitations and opportunities of integrating artificial intelligence into breast cancer clinical practice.

Improving breast cancer diagnostics with deep learning for MRI

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) has a high sensitivity in detecting breast cancer but often leads to unnecessary biopsies and patient workup. We used a deep learning (DL) system to improve the overall accuracy of breast cancer diagnosis and personalize management of patients undergoing DCE-MRI. On the internal test set (n = 3936 exams), our system achieved an area under the receiver operating characteristic curve (AUROC) of 0.92 (95% CI: 0.92 to 0.93). In a retrospective reader study, there was no statistically significant difference (P = 0.19) between five board-certified breast radiologists and the DL system (mean ΔAUROC, +0.04 in favor of the DL system). Radiologists' performance improved when their predictions were averaged with DL's predictions [mean ΔAUPRC (area under the precision-recall curve), +0.07]. We demonstrated the generalizability of the DL system using multiple datasets from Poland and the United States. An additional reader study on a Polish dataset showed that the DL system was as robust to distribution shift as radiologists. In subgroup analysis, we observed consistent results across different cancer subtypes and patient demographics. Using decision curve analysis, we showed that the DL system can reduce unnecessary biopsies in the range of clinically relevant risk thresholds. This would lead to avoiding biopsies yielding benign results in up to 20% of all patients with BI-RADS category 4 lesions. Last, we performed an error analysis, investigating situations where DL predictions were mostly incorrect. This exploratory work creates a foundation for deployment and prospective analysis of DL-based models for breast MRI.

Using deep learning to safely exclude lesions with only ultrafast breast MRI to shorten acquisition and reading time

OBJECTIVES

To investigate the feasibility of automatically identifying normal scans in ultrafast breast MRI with artificial intelligence (AI) to increase efficiency and reduce workload.

METHODS

In this retrospective analysis, 837 breast MRI examinations performed on 438 women from April 2016 to October 2019 were included. The left and right breasts in each examination were labelled normal (without suspicious lesions) or abnormal (with suspicious lesions) based on final interpretation. Maximum intensity projection (MIP) images of each breast were then used to train a deep learning model. A high sensitivity threshold was calculated based on the detection trade - off (DET) curve on the validation set. The performance of the model was evaluated by receiver operating characteristic analysis of the independent test set. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) with the high sensitivity threshold were calculated.

RESULTS

The independent test set consisted of 178 examinations of 149 patients (mean age, 44 years ± 14 [standard deviation]). The trained model achieved an AUC of 0.81 (95% CI: 0.75-0.88) on the independent test set. Applying a threshold of 0.25 yielded a sensitivity of 98% (95% CI: 90%; 100%), an NPV of 98% (95% CI: 89%; 100%), a workload reduction of 15.7%, and a scan time reduction of 16.6%.

CONCLUSION

This deep learning model has a high potential to help identify normal scans in ultrafast breast MRI and thereby reduce radiologists' workload and scan time.

KEY POINTS

• Deep learning in TWIST may eliminate the necessity of additional sequences for identifying normal breasts during MRI screening. • Workload and scanning time reductions of 15.7% and 16.6%, respectively, could be achieved with the cost of 1 (1 of 55) false negative prediction.