Overview

Schema Version

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

Name

Semiautonomous Deep Learning Mammography Rule-out Study Datasets

Link

https://pubmed.ncbi.nlm.nih.gov/38597785/

Indexing

Keywords: Artificial Intelligence, Semiautonomous Deep Learning, Breast Cancer, Screening Mammography, Rule-out device, Full-field digital mammography

Content: BR, SQ

RadLex: RID10357

SNOMED: 254837009

Author(s)

Stefano Pedemonte, PhD

Trevor Tsue, MS

Brent Mombourquette, MS

Yen Nhi Truong Vu, MS

Thomas Matthews, PhD

Rodrigo Morales Hoil, BS

Meet Shah, MS

Nikita Ghare, BS

Naomi Zingman-Daniels, MHSc

Susan Holley, MD, PhD

Catherine M. Appleton, MD

Jason Su, MS

Richard L. Wahl, MD

Organization(s)

Whiterabbit.ai

Onsite Women's Health

SSM Health

Mallinckrodt Institute of Radiology, Washington University School of Medicine

Royal Surrey NHS Foundation Trust (OPTIMAM)

Cancer Research UK (OPTIMAM funding)

License

Text: CC BY 4.0

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

Contact

Corresponding author: Trevor Tsue

Funding

Supported by Whiterabbit.ai. Washington University in St. Louis has equity interests in Whiterabbit.ai and may receive royalties per agreement to develop the technology.

Ethical review

IRB-approved use of anonymized data; informed consent waived; handled per HIPAA.

Comments

Retrospective anonymized 2D full-field digital mammography datasets from two U.S. institutions and one U.K. institution (OPTIMAM) used to train and evaluate a breast cancer rule-out AI system.

Date

Published: 2024-04-10

References

[1] Pedemonte S, Tsue T, Mombourquette B, et al.. "A Semiautonomous Deep Learning System to Reduce False Positives in Screening Mammography". Radiology: Artificial Intelligence. 2024-05-01. doi:10.1148/ryai.230033. PMID: 38597785. PMCID: PMC11140506.

[2] Halling-Brown MD, Warren LM, Ward D, et al.. "OPTIMAM Mammography Image Database: A Large-Scale Resource of Mammography Images and Clinical Data". Radiology: Artificial Intelligence. 2020-01-01. PMID: 33937853. PMCID: PMC8082293. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8082293/

Dataset

Motivation

Evaluate an AI rule-out device to reduce false positives and workload in screening mammography while maintaining cancer detection.

Sampling

Inclusion/exclusion per Figure 1 in article; N/S/D labels required ≥2-year follow-up and no biopsies; interval cancer windows: U.S. 12 months (also 6, 24 used), U.K. up to 36 months.

Partitioning scheme

U.S. dataset 1 and U.K. dataset 3 split by patient into 80% training, 10% validation, 10% test; U.S. dataset 2 fully held out for testing.

Missing information

Image resolution and series/image counts not reported.

Relationships between instances

Labels assigned per breast and propagated to examinations by maximum outcome; multiview and prior examinations used by the model.

External data

U.K. dataset derived from the OPTIMAM Mammography Image Database.

Confidentiality

Retrospective anonymized patient data; HIPAA compliant.

Re-identification

Data anonymized; labels based on BI-RADS assessments, reader opinions, and pathology with follow-up requirements.

Sensitive data

Breast imaging of female patients; de-identified prior to analysis.