Overview

Schema Version

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

Name

PriorNet dataset: Digital breast tomosynthesis screening dataset with 1-year priors (2016–2020)

Link

https://dx.doi.org/10.1148/ryai.230391

Indexing

Keywords: Digital Breast Tomosynthesis, Computer-aided Detection, Breast Cancer, Deep Learning, Temporal comparison, Prior examination, Hologic

Content: BR

RadLex: RID10359, RID45682, RID45702, RID10357

SNOMED: 254837009

Author(s)

Yinhao Ren

Zisheng Liang

Jun Ge

Xiaoming Xu

Jonathan Go

Derek L. Nguyen

Joseph Y. Lo

Lars J. Grimm

Organization(s)

Duke University

iCAD Inc

Contact

Corresponding author: Yinhao Ren (yinhao.ren@duke.edu)

Funding

Supported by iCAD under sponsored research collaboration agreement (SPS#264557).

Ethical review

HIPAA-compliant retrospective study; sponsor’s institutional review board approved; informed consent requirement waived.

Comments

Retrospective multi-site DBT screening dataset including current and 1-year-prior examinations; positive cases were histopathology-proven cancers; negatives were BI-RADS 1–2 or benign histopathology.

Date

Published: 2024-08-14

References

[1] Ren Y, Liang Z, Ge J, Xu X, Go J, Nguyen DL, Lo JY, Grimm LJ. "Improving Computer-aided Detection for Digital Breast Tomosynthesis by Incorporating Temporal Change". Radiology: Artificial Intelligence. 2024-08-14. doi:10.1148/ryai.230391. PMID: 39140867. PMCID: PMC11427939.

Dataset

Motivation

To incorporate temporal growth information from prior-year DBT examinations to improve cancer lesion detection.

Sampling

Retrospective collection of consecutive screening DBT cases from eight sites in the US and Europe (2016–2020), all acquired on Hologic systems.

Partitioning scheme

Data from seven sites split by patient into training and validation; the eighth site reserved for external testing.

Missing information

Exact image file formats, raw image resolutions, and detailed site identifiers are not reported.

Relationships between instances

Each case consists of a pair of examinations (current and 1-year-prior), with multiple DBT views (CC and MLO) per breast; lesion-level annotations for current exams and prior locations provided for training.

Confidentiality

HIPAA compliant; IRB approved; consent waived.

Sensitive data

Contains clinical imaging data with lesion annotations; PHI handling not detailed beyond HIPAA compliance.