Overview

Schema Version

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

Name

ADMANI: Annotated Digital Mammograms and Associated Non-Image Datasets

Link

https://dx.doi.org/10.1148/ryai.220072

Indexing

Keywords: ADMANI, Mammography, Screening, Breast cancer detection, Interval cancer, Convolutional Neural Network (CNN), Digital mammography

Content: BR

RadLex: RID45682, RID10519, RID50114, RID10357

SNOMED: 254837009, 444589003

Author(s)

Helen M. L. Frazer

Jennifer S. N. Tang

Michael S. Elliott

Katrina M. Kunicki

Brendan Hill

Ravishankar Karthik

Chun Fung Kwok

Carlos A. Peña-Solorzano

Yuanhong Chen

Chong Wang

Osamah Al-Qershi

Samantha K. Fox

Shuai Li

Enes Makalic

Tuong L. Nguyen

Daniel F. Schmidt

Prabhathi Basnayake Ralalage

Jocelyn F. Lippey

Peter Brotchie

John L. Hopper

Gustavo Carneiro

Davis J. McCarthy

Organization(s)

St Vincent's BreastScreen, St Vincent's Hospital Melbourne

BreastScreen Victoria

St Vincent's Institute of Medical Research

University of Melbourne

University of Adelaide

Centre for Epidemiology & Biostatistics, Melbourne School of Population and Global Health

Monash University

Melbourne Integrative Genomics

Version

1.0

License

Text: CC BY 4.0

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

Contact

ua.gro.ahvs@rezarF.neleH

Funding

Australian Government MRFF Applied AI Research in Health grant (no. MRFAI000090) supporting the BRAIx program; Ramaciotti Health Investment Grant (Ramaciotti Foundation, Australia); in-kind contributions from St Vincent's Institute of Medical Research, St Vincent's Hospital Melbourne, BreastScreen Victoria, University of Melbourne, and University of Adelaide.

Ethical review

Use governed under BRAIx Multi-Institutional Agreement with Human Research Ethics Committee approvals LNR/18/SVHM/162 and LNR/19/SVHM/123; de-identified data with unique identifiers; consent obtained at screening registration for research use.

Date

Updated: 2022-12-06

Published: 2022-12-21

Created: 2022-04-16

References

[1] Frazer HML, Tang JSN, Elliott MS, Kunicki KM, Hill B, Karthik R, Kwok CF, Peña-Solorzano CA, Chen Y, Wang C, Al-Qershi O, Fox SK, Li S, Makalic E, Nguyen TL, Schmidt DF, Basnayake Ralalage P, Lippey JF, Brotchie P, Hopper JL, Carneiro G, McCarthy DJ. "ADMANI: Annotated Digital Mammograms and Associated Non-Image Datasets". Radiology: Artificial Intelligence. 2023 Mar;5(2):e220072.. 2023-03-01. doi:10.1148/ryai.220072. PMID: 37035431. PMCID: PMC10077091.

Model

Availability

Large-scale, multi-center, clinically curated datasets (ADMANI1–3). A subset of 40,000 images from 10,000 episodes provided for the RSNA Mammography Breast Cancer Detection AI Challenge launching Nov 28, 2022; the challenge training dataset will be made public and remain available to researchers. Future broader availability under development (funding and governance in progress).

Clinical benefit

Supports development and evaluation of AI algorithms for breast cancer detection in population screening and for risk-based screening research.

Clinical workflow phase

Dataset enables evaluation at different stages of the screening pathway (triage, reader support, potential reader replacement, decision support).

Indications for use

To develop, train, and evaluate AI systems for detecting screen-detected and interval breast cancers in a population-based mammographic screening setting; intended population includes women participating in BreastScreen Victoria (primarily ages 50–74, available from age 40).

Input

For-presentation 2D full-field digital mammograms (bilateral CC and MLO) with associated non-image data: patient demographics and risk, symptoms, radiologist reading data (lesion side/grade, annotations), and histopathology (surgical specimen results, subtype).

Limitations

Current datasets do not include digital breast tomosynthesis or breast ultrasound images; more research needed to characterize interval cancers detectable by AI vs de novo during screening interval; non-transformed source data under license with the state screening program; broader dataset availability pending funding and governance.

Use

Intended: Detection, Detection and diagnosis

User

Intended: Other, Researcher