Overview

Schema Version

https://atlas.rsna.org/schemas/2025-11/model.json

Name

iBRISK: intelligent-augmented breast cancer risk calculator

Link

https://dx.doi.org/10.1148/ryai.220259

Indexing

Keywords: Mammography, Breast, Oncology, Biopsy/Needle Aspiration, Radiomics, Precision Mammography, AI-augmented Biopsy Decision Support Tool, Breast Cancer Risk Calculator, BI-RADS 4, Risk stratification, Overbiopsy reduction, Probability of Malignancy (POM), PPV3

Content: BR, OI

RadLex: RID34642, RID36030, RID39055

SNOMED: 254837009

Author(s)

Chika F. Ezeana

Tiancheng He

Tejal A. Patel

Virginia Kaklamani

Maryam Elmi

Erika Brigmon

Pamela M. Otto

Kenneth A. Kist

Heather Speck

Lin Wang

Joe Ensor

Ya-Chen T. Shih

Bumyang Kim

I-Wen Pan

Adam L. Cohen

Kristen Kelley

David Spak

Wei T. Yang

Jenny C. Chang

Stephen T. C. Wong

Organization(s)

Houston Methodist Neal Cancer Center, Houston Methodist Hospital

University of Texas MD Anderson Cancer Center

University of Texas Health Science Center San Antonio

University of the Incarnate Word School of Osteopathic Medicine

Huntsman Cancer Institute, University of Utah

Weill Cornell Medicine / Houston Methodist Hospital

Version

1.0

License

Text: CC BY 4.0

URL: https://creativecommons.org/licenses/by/4.0/

Contact

Corresponding author: Stephen T. C. Wong; email: gro.tsidohtemnotsuoh@gnowts

Funding

Supported by the Ting Tsung & Wei Fong Chao Family Foundation, the John S. Dunn Research Foundation, the Breast Cancer Research Foundation, and NIH/NCI grant no. 1R01CA251710.

Ethical review

IRB-approved, HIPAA-compliant multicenter retrospective study with waivers of informed consent at participating institutions.

Date

Published: 2023-08-09

Created: 2022-11-25

References

[1] Ezeana CF, He T, Patel TA, et al.. "A Deep Learning Decision Support Tool to Improve Risk Stratification and Reduce Unnecessary Biopsies in BI-RADS 4 Mammograms". Radiology: Artificial Intelligence. 2023;5(6):e220259.. 2023-08-09. doi:10.1148/ryai.220259. PMID: 38074778. PMCID: PMC10698614.

[2] He T, Puppala M, Ezeana CF, et al.. "A deep learning-based decision support tool for precision risk assessment of breast cancer". JCO Clinical Cancer Informatics. 2019;3:1-12.. 2019-01-01. PMID: 31141423. PMCID: PMC10445790. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10445790/

Model

Architecture

Deep learning model integrating mammographic descriptors and clinical risk factors (details provided in Appendix S1 of the article).

Availability

Evaluated retrospectively; authors state the iBRISK calculator will be published as an online, open-access, noncommercial interface. Data used are available from the corresponding author by request.

Clinical benefit

Improves risk stratification of BI-RADS 4 lesions, enabling safe downgrading of low/moderate-risk cases to reduce unnecessary biopsies and associated costs and distress.

Clinical workflow phase

Clinical decision support systems; biopsy decision support adjunct following diagnostic mammography with BI-RADS 4 assessment.

Decision threshold

Trichotomized thresholds: low POM < 0.40; moderate POM 0.40–0.55; high POM > 0.55.

Degree of automation

Decision support (assists clinicians; does not automate final diagnosis or replace BI-RADS).

Indications for use

Assessment of probability of malignancy in women with BI-RADS category 4 mammographic lesions in diagnostic settings across diverse clinical sites; intended for use as an adjunct to BI-RADS to inform biopsy decision-making.

Input

Twenty variables per patient comprising clinical risk factors (demographics, metabolic factors, history, physical signs) and mammographic descriptors (density, mass, calcification features, asymmetry, architectural distortion).

Instructions

After a BI-RADS 4 designation, enter the 20 clinical and mammographic descriptor features into the iBRISK calculator to obtain a probability of malignancy (0–1) and decision support recommendations for biopsy triage.

Limitations

Retrospective multicenter study largely limited to three Texas institutions; requires curated inputs with up to three missing features tolerated; substitutes missing values (unknown/average), which can affect accuracy; variability and inconsistency in reporting of calcification features; model built/refined on data from a single health system with external testing at three sites; not a definitive diagnostic tool; requires structured reporting for optimal performance.

Output

CDEs: RDE65, RDE1586, RDE2077

Description: Probability of malignancy (POM) score between 0 and 1 with categorical risk (low/moderate/high) and biopsy decision support recommendation.

Recommendation

Use as an adjunct to BI-RADS to triage BI-RADS 4 lesions: consider avoiding biopsy in low and moderate POM groups (potentially up to 50% of cases) and manage high POM similar to BI-RADS 5.

Reproducibility

Missing feature analysis (MDACC, n=1424) showed slight accuracy declines with each additional missing feature; statistically significant drop when the fourth feature was removed, indicating robustness with up to three missing inputs.

Use

Intended: Decision support

Out-of-scope: Diagnosis

Excluded: Diagnosis

User

Intended: Referring physician, Radiologist, Subspecialist diagnostic radiologist