Overview

Schema Version

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

Name

FIND Validation Platform chest radiography datasets for independent CAD evaluation in tuberculosis

Link

https://www.finddx.org

Indexing

Keywords: tuberculosis, chest radiography, computer-aided detection, deep learning, screening, triage, low- and middle-income countries, independent validation, FIND validation platform

Content: CH

RadLex: RID28818, RID45946, RID45702, RID29116, RID1243

SNOMED: 154283005

Author(s)

FIND – The Global Alliance for Diagnostics

Matthew Arentz

Morten Ruhwald

Organization(s)

FIND – The Global Alliance for Diagnostics

World Health Organization (WHO)

International Telecommunication Union (ITU)

World Intellectual Property Organization (WIPO)

Contact

FIND (WHO Collaborating Center) – https://www.finddx.org

Funding

Data collection and development of the FIND validation platform supported by the German Ministry for Education and Research (BMBF) through KfW.

Ethical review

Data curation adheres to international standards for ethical research; datasets are drawn from published cohorts for transparency.

Comments

The article describes curated, high-quality chest radiography datasets assembled by FIND for independent, iterative evaluations of CAD tools for tuberculosis. Current datasets encompass almost 20,000 chest radiographs from multiple countries in five WHO regions; data are drawn from published cohorts and adhere to international standards for ethical research; data are not shared with developers.

Date

Published: 2024-01-10

References

[1] Hwang EJ, Jeong WG, David PM, Arentz M, Ruhwald M, Yoon SH. "AI for Detection of Tuberculosis: Implications for Global Health". Radiology: Artificial Intelligence. 2024-01-10. doi:10.1148/ryai.230327. PMID: 38197795. PMCID: PMC10982823.

[2] FIND. "Report from the FIND meeting on March 8, 2023: WHO Recommendations and Developments in the Prequalification of Chest Radiography and Computer-Aided Detection (CXR-CAD) for Tuberculosis". FIND. 2023-01-01. Available from: https://www.finddx.org/wp-content/uploads/2023/03/20220324_rep_crx_cad_tb_FV_EN.pdf

Dataset

Motivation

Enable rapid, manufacturer-independent evaluations of CAD tools for TB across diverse populations and use cases to inform WHO guidance and country adoption.

Sampling

Collected from multiple countries across five WHO regions; specific sampling strategies are not detailed.

Missing information

No per-site counts, demographics, or image formats are specified in the article.

External data

Datasets are drawn from published cohorts to maximize transparency; not shared with developers.

Confidentiality

Clinical imaging and associated metadata from published cohorts; evaluated within a secure platform; developer access removed during evaluations.

Re-identification

Not described; data use follows ethical standards; data are not shared with developers and are evaluated within a secure environment.

Sensitive data

Health imaging and clinical data related to TB diagnosis.