BI-RADS decoded: detailed guidance on potentially confusing issues

Mammographic features and risk of breast cancer death among women with invasive screen-detected cancer in BreastScreen Norway 1996-2020

OBJECTIVES

We explored associations between mammographic features and risk of breast cancer death among women with small (<15 mm) and large (≥15 mm) invasive screen-detected breast cancer.

METHODS

We included data from 17,614 women diagnosed with invasive breast cancer as a result of participation in BreastScreen Norway, 1996-2020. Data on mammographic features (mass, spiculated mass, architectural distortion, asymmetric density, density with calcification and calcification alone), tumour diameter and cause of death was obtained from the Cancer Registry of Norway. Cox regression was used to estimate hazard ratios (HR) with 95% confidence intervals (CI) for breast cancer death by mammographic features using spiculated mass as reference, adjusting for age, tumour diameter and lymph node status. All analyses were dichotomised by tumour diameter (small versus large).

RESULTS

Mean age at diagnosis was 60.8 (standard deviation, SD=5.8) for 10,160 women with small tumours and 60.0 (SD=5.8) years for 7454 women with large tumours. The number of breast cancer deaths was 299 and 634, respectively. Mean time from diagnosis to death was 8.7 (SD=5.0) years for women with small tumours and 7.2 (4.6) years for women with large tumours. Using spiculated mass as reference, adjusted HR for breast cancer death among women with small tumours was 2.48 (95% CI 1.67-3.68) for calcification alone, while HR for women with large tumours was 1.30 (95% CI 1.02-1.66) for density with calcification.

CONCLUSIONS

Small screen-detected invasive cancers presenting as calcification and large screen-detected cancers presenting as density with calcification were associated with the highest risk of breast cancer death.

CLINICAL RELEVANCE STATEMENT

Small tumours (<15 mm) presented as calcification alone and large tumours (≥ 15 mm) presented as density with calcification were associated with the highest risk of breast cancer death among women with screen-detected invasive breast cancer diagnosed 1996-2020.

KEY POINTS

• Women diagnosed with invasive screen-detected breast cancer 1996-2020 were analysed. • Small screen-detected cancers presenting as calcification alone resulted in the highest risk of breast cancer death. • Large screen-detected cancers presenting as density with calcification resulted in the highest risk of breast cancer death.

Performance of an Artificial Intelligence System for Breast Cancer Detection on Screening Mammograms from BreastScreen Norway

Purpose To explore the stand-alone breast cancer detection performance, at different risk score thresholds, of a commercially available artificial intelligence (AI) system. Materials and Methods This retrospective study included information from 661 695 digital mammographic examinations performed among 242 629 female individuals screened as a part of BreastScreen Norway, 2004-2018. The study sample included 3807 screen-detected cancers and 1110 interval breast cancers. A continuous examination-level risk score by the AI system was used to measure performance as the area under the receiver operating characteristic curve (AUC) with 95% CIs and cancer detection at different AI risk score thresholds. Results The AUC of the AI system was 0.93 (95% CI: 0.92, 0.93) for screen-detected cancers and interval breast cancers combined and 0.97 (95% CI: 0.97, 0.97) for screen-detected cancers. In a setting where 10% of the examinations with the highest AI risk scores were defined as positive and 90% with the lowest scores as negative, 92.0% (3502 of 3807) of the screen-detected cancers and 44.6% (495 of 1110) of the interval breast cancers were identified with AI. In this scenario, 68.5% (10 987 of 16 040) of false-positive screening results (negative recall assessment) were considered negative by AI. When 50% was used as the cutoff, 99.3% (3781 of 3807) of the screen-detected cancers and 85.2% (946 of 1110) of the interval breast cancers were identified as positive by AI, whereas 17.0% (2725 of 16 040) of the false-positive results were considered negative. Conclusion The AI system showed high performance in detecting breast cancers within 2 years of screening mammography and a potential for use to triage low-risk mammograms to reduce radiologist workload. Keywords: Mammography, Breast, Screening, Convolutional Neural Network (CNN), Deep Learning Algorithms Supplemental material is available for this article. © RSNA, 2024 See also commentary by Bahl and Do in this issue.

The Longitudinal Imaging Tracker (BrICS-LIT):A Cloud Platform for Monitoring Treatment Response in Glioblastoma Patients

Glioblastoma is a common and aggressive form of brain cancer affecting up to 20,000 new patients in the US annually. Despite rigorous therapies, current median survival is only 15-20 months. Patients who complete initial treatment undergo follow-up imaging at routine intervals to assess for tumor recurrence. Imaging is a central part of brain tumor management, but MRI findings in patients with brain tumor can be challenging to interpret and are further confounded by interpretation variability. Disease-specific structured reporting attempts to reduce variability in imaging results by implementing well-defined imaging criteria and standardized language. The Brain Tumor Reporting and Data System (BT-RADS) is one such framework streamlined for clinical workflows and includes quantitative criteria for more objective evaluation of follow-up imaging. To facilitate accurate and objective monitoring of patients during the follow-up period, we developed a cloud platform, the Brain Imaging Collaborative Suite's Longitudinal Imaging Tracker (BrICS-LIT). BrICS-LIT uses semiautomated tumor segmentation algorithms of both T2-weighted FLAIR and contrast-enhanced T1-weighted MRI to assist clinicians in quantitative assessment of brain tumors. The LIT platform can ultimately guide clinical decision-making for patients with glioblastoma by providing quantitative metrics for BT-RADS scoring. Further, this platform has the potential to increase objectivity when measuring efficacy of novel therapies for patients with brain tumor during their follow-up. Therefore, LIT will be used to track patients in a dose-escalated clinical trial, where spectroscopic MRI has been used to guide radiation therapy (Clinicaltrials.gov NCT03137888), and compare patients to a control group that received standard of care.

Risk of breast cancer two years after a benign biopsy depends on the mammographic feature prompting recall

OBJECTIVE

We aimed to explore whether the type of mammographic feature prompting a false-positive recall (FPR) during mammography screening influences the risk and timing of breast cancer diagnosis, particularly if assessed with invasive procedures.

STUDY DESIGN

We included information on women screened and recalled for further assessment in Spain between 1994 and 2015, with follow-up until 2017, categorizing FPRs by the assessment (noninvasive or invasive) and mammographic feature prompting the recall.

MAIN OUTCOME MEASURES

Breast cancer rates in the first two years after FPR (first period) and after two years (second period).

RESULTS

The study included 99,825 women with FPRs. In both periods, the breast cancer rate was higher in the invasive assessment group than in the noninvasive group (first period 12 ‰ vs 1.9 ‰, p < 0.001; second period 4.4‰ vs 3.1‰, p < 0.001). During the first period, the invasive assessment group showed diverse breast cancer rates for each type of mammographic feature, with a higher rate for asymmetric density (31.9‰). When the second period was compared with the first, the breast cancer rate decreased in the invasive assessment group (from 12‰ to 4.4‰, p < 0.001) and increased in the noninvasive assessment group (from 1.9‰ to 3.1‰, p < 0.001).

CONCLUSION

In the context of mammography screening, the risk of breast cancer diagnosis during the first two years after FPR was particularly high for women undergoing invasive assessment; importantly, the risk was modified by type of mammographic feature prompting the recall. This information could help to individualize follow-up after exclusion of malignancy.

[Meta analysis of the use of Bayesian networks in breast cancer diagnosis]

The aim of this study was to determine the accuracy of Bayesian networks in supporting breast cancer diagnoses. Systematic review and meta-analysis were carried out, including articles and papers published between January 1990 and March 2013. We included prospective and retrospective cross-sectional studies of the accuracy of diagnoses of breast lesions (target conditions) made using Bayesian networks (index test). Four primary studies that included 1,223 breast lesions were analyzed, 89.52% (444/496) of the breast cancer cases and 6.33% (46/727) of the benign lesions were positive based on the Bayesian network analysis. The area under the curve (AUC) for the summary receiver operating characteristic curve (SROC) was 0.97, with a Q* value of 0.92. Using Bayesian networks to diagnose malignant lesions increased the pretest probability of a true positive from 40.03% to 90.05% and decreased the probability of a false negative to 6.44%. Therefore, our results demonstrated that Bayesian networks provide an accurate and non-invasive method to support breast cancer diagnosis.

Developing a comprehensive database management system for organization and evaluation of mammography datasets

We aimed to design and develop a comprehensive mammography database system (CMDB) to collect clinical datasets for outcome assessment and development of decision support tools. A Health Insurance Portability and Accountability Act (HIPAA) compliant CMDB was created to store multi-relational datasets of demographic risk factors and mammogram results using the Breast Imaging Reporting and Data System (BI-RADS) lexicon. The CMDB collected both biopsy pathology outcomes, in a breast pathology lexicon compiled by extending BI-RADS, and our institutional breast cancer registry. The audit results derived from the CMDB were in accordance with Mammography Quality Standards Act (MQSA) audits and national benchmarks. The CMDB has managed the challenges of multi-level organization demanded by the complexity of mammography practice and lexicon development in pathology. We foresee that the CMDB will be useful for efficient quality assurance audits and development of decision support tools to improve breast cancer diagnosis. Our procedure of developing the CMDB provides a framework to build a detailed data repository for breast imaging quality control and research, which has the potential to augment existing resources.

Predicting visual semantic descriptive terms from radiological image data: preliminary results with liver lesions in CT

We describe a framework to model visual semantics of liver lesions in CT images in order to predict the visual semantic terms (VST) reported by radiologists in describing these lesions. Computational models of VST are learned from image data using linear combinations of high-order steerable Riesz wavelets and support vector machines (SVM). In a first step, these models are used to predict the presence of each semantic term that describes liver lesions. In a second step, the distances between all VST models are calculated to establish a nonhierarchical computationally-derived ontology of VST containing inter-term synonymy and complementarity. A preliminary evaluation of the proposed framework was carried out using 74 liver lesions annotated with a set of 18 VSTs from the RadLex ontology. A leave-one-patient-out cross-validation resulted in an average area under the ROC curve of 0.853 for predicting the presence of each VST. The proposed framework is expected to foster human-computer synergies for the interpretation of radiological images while using rotation-covariant computational models of VSTs to 1) quantify their local likelihood and 2) explicitly link them with pixel-based image content in the context of a given imaging domain.

Evaluation of the PI-RADS scoring system for mpMRI of the prostate: a whole-mount step-section analysis

PURPOSE

Evaluation of the Prostate Imaging Reporting and Data System (PI-RADS) scoring system for classifying multi-parametric magnetic resonance imaging findings of the prostate using whole-mount step-section slides as reference standard.

MATERIALS AND METHODS

Prospective inclusion of 50 consecutive patients with biopsy-proven prostate cancer (PCa). All patients received a multi-parametric MRI of the prostate, consisting of T2-weighted, diffusion-weighted, and dynamic contrast-enhanced MRI. After prostatectomy, all prostates were prepared as whole-mount step-section slides. For each patient, six lesions were predefined on whole-mount step-sections according to a distinct scheme and the corresponding regions were identified on MRI. Each lesion then was scored on MRI according to PI-RADS by an experienced blinded uro-radiologist and compared with histopathological findings.

RESULTS

PCa received significant (p < 0.01) higher overall PI-RADS scores (4.10 ± 0.75) compared with benign changes (2.00 ± 0.74). In the peripheral zone, each single modality score showed good diagnostic accuracy for PCa detection (area under the curve [AUC] > 0.90). When combining all single modality scores, an even higher discriminative ability of PCa detection (AUC = 0.97, 95 % CI 0.95-0.99) could be achieved. In contrast, in the transitional zone, dynamic contrast-enhanced MRI (DCE) showed very low diagnostic accuracy (AUC = 0.60). Regarding tumor malignancy, no high-grade PCa (Gleason >7a) was present at PI-RADS scores <4 and no Gleason 6 PCa at a PI-RADS score of 5.

CONCLUSION

The PI-RADS scoring system showed good diagnostic accuracy: Only PI-RADS 4 and 5 showed high-grade PCa. However, it seems necessary to revise the PI-RADS scoring system concerning DCE in the transitional zone.

Performance of double reading mammography in an Iranian population and its effect on patient outcome

Background

Considering the importance and responsibility of reporting mammography and the necessity to notice details with a high degree of precision, double reading mammography has been introduced and recommended.

Objectives

This study aimed to assess the performance of double reading of mammograms and its effect on patient outcomes.

Patients and Methods

Throughout this cross sectional study, 1284 digitized mammographic views of 642 breasts which belonged to 339 women (of which 303 were bilateral and 36 were unilateral mammographies) were enrolled. Two independent radiologists interpreted these mammograms and BI-RADS categories of both reports were compared. Discordant results were determined and assumed significant if they were in the positive (BI-RADS 0, 4, 5) versus negative (BI-RADS 1, 2, 3) groups and then significant discordant cases were followed up to determine benign versus malignant final diagnosis. The recall rate was calculated for each reader. Inter-observer agreement in breast density was determined by Kappa test.

Results

Readers had consensus on BI-RADS categories in 459 breasts (71%), but diverse categories were used for 183 breasts (29%), including 132 significant and 51 non-significant discrepancies. According to weighted Kappa test, agreement between two readers in positive or negative reports was 0.78 (95% CI=0.73-0.83) and in parenchymal density, it was 0.73 (95% CI=0.7-0.77). Most of the discrepancies were between category zero versus categories 1 and 2 (63.4%). The recall rate was 36% for the first and 44% for the second reader. Among 132 significant discordant results, one case had the final diagnosis of malignancy and the others had benign or negative diagnosis. There was 0.2% increase in cancer detection rate by double reading.

Conclusion

This study shows no significant improvement in the cancer detection rate by double reading; however, a lower recall rate could be a more helpful consequence.

Multidisciplinary care of patients with early-stage breast cancer

There is a compelling need for close coordination and integration of multiple specialties in the management of patients with early-stage breast cancer. Optimal patient care and outcomes depend on the sequential and often simultaneous participation and dialogue between specialists in imaging, pathologic and molecular diagnostic and prognostic stratification, and the therapeutic specialties of surgery, radiation oncology, and medical oncology. These are but a few of the various disciplines needed to provide modern, sophisticated management. The essential role for coordinated involvement of the entire health care team in optimal management of patients with early-stage breast cancer is likely to increase further.

An Effective Way to Solve Equivocal Mammography Findings: The Rolled Views

SUMMARY: BACKGROUND: The aim of this study was to investigate the efficacy of the rolled views taken in craniocaudal (CC) and mediolateral oblique (MLO) projections in solving equivocal mammography findings. PATIENTS AND METHODS: The rolled views were taken by changing the positioning of the breast but not the obliquity of the X-ray beams. The breast was rolled medially or laterally in the rolled CC view, and inferiorly or superiorly in the rolled MLO view to separate overlapping structures from each other. RESULTS: We evaluated equivocal findings in 87 asymptomatic women undergoing either CC (n = 48, 55%) or MLO (n = 39, 45%) rolled views between 2001 and 2008. The rolled views were helpful in solving equivocal mammographic findings and making proper decisions on management in 85 of the 87 (97.7%) women. This technique was used for breast asymmetries in 55 of the 87 (63.2%) women, and was sufficient to directly show summation artifacts in 59 of 79 (74.6%) women. The rolled views revealed 4 intramammary lymph nodes, 2 circumscribed masses out of 6 obscured masses, 7 summation artifacts, and 2 circumscribed masses out of 9 questionable masses. CONCLUSIONS: The rolled view is an effective method of differentiating summation artifacts from real lesions on mammography in both the CC and the MLO view.

Quantitative methodology using CT for predicting survival in patients with metastatic colorectal carcinoma: a pilot study

OBJECTIVE

To develop a methodology which quantifies multiple changing lesion features resulting in an optimized computed tomography (CT) response score (CRS) for prediction of overall survival (OS) in response to treatment for metastatic colorectal carcinoma (MCRC).

SUBJECTS AND METHODS

This Health Insurance Portability and Accountability Act-compliant, institutional review board-approved retrospective study evaluated multiple changing imaging findings and their correlation with OS with a new methodology comparing the baseline and first post-treatment CT scans in 38 MCRC patients on last-line chemotherapy (cetuximab and irinotecan). Tumor size/enhancement changes and interval development of new lesions were quantified with either Likert-type scales (all parameters) or Response Evaluation Criteria in Solid Tumors (RECIST) (size change only). The most predictive parameters for OS were used to generate the CRS with an overall range of -3 (complete disappearance) to +2 (definite tumor increase). The Cox Hazard Ratio was used to assess prediction of survival. Reader agreement was evaluated by the kappa statistic.

RESULTS

Tumor size was the best predictor of OS using the Likert-type scale or RECIST. The CRS was not improved combining size change with other parameters. Use of the Likert-type scale resulted in predicting OS with a Cox hazard ratio of 1.697 (P=.0004) and good agreement (kappa=0.73, 95% CI=0.41-1.10) between observers with no significant difference using RECIST.

CONCLUSION

The methodology produces a CRS for MCRC predicting OS resulting from therapy which expands standard RECIST guidelines to allow critical evaluation of multiple additional imaging parameters. Size change alone was found to be the best parameter of those considered in terms of maximizing agreement and prediction of OS.

The ACR BI-RADS experience: learning from history

The Breast Imaging Reporting and Data System (BI-RADS) initiative, instituted by the ACR, was begun in the late 1980s to address a lack of standardization and uniformity in mammography practice reporting. An important component of the BI-RADS initiative is the lexicon, a dictionary of descriptors of specific imaging features. The BI-RADS lexicon has always been data driven, using descriptors that previously had been shown in the literature to be predictive of benign and malignant disease. Once established, the BI-RADS lexicon provided new opportunities for quality assurance, communication, research, and improved patient care. The history of this lexicon illustrates a series of challenges and instructive successes that provide a valuable guide for other groups that aspire to develop similar lexicons in the future.

Breast tomosynthesis in clinical practice: initial results

PURPOSE

The purpose of this study was to assess the potential value of tomosynthesis in women with an abnormal screening mammogram or with clinical symptoms. Mammography and tomosynthesis investigations of 513 woman with an abnormal screening mammogram or with clinical symptoms were prospectively classified according to the ACR BI-RADS criteria. Sensitivity and specificity of both techniques for the detection of cancer were calculated. In 112 newly detected cancers, tomosynthesis and mammography were each false-negative in 8 cases (7%). In the false-negative mammography cases, the tumor was detected with ultrasound (n = 4), MRI (n = 2), by recall after breast tomosynthesis interpretation (n = 1), and after prophylactic mastectomy (n = 1). Combining the results of mammography and tomosynthesis detected 109 cancers. Therefore in three patients, both mammography and tomosynthesis missed the carcinoma. The sensitivity of both techniques for the detection of breast cancer was 92.9%, and the specificity of mammography and tomosynthesis was 86.1 and 84.4%, respectively. Tomosynthesis can be used as an additional technique to mammography in patients referred with an abnormal screening mammogram or with clinical symptoms. Additional lesions detected by tomosynthesis, however, are also likely to be detected by other techniques used in the clinical work-up of these patients.