qure_logo.svg

Published 01 May 2023

The rise of artificial intelligence reading of chest X-rays for enhanced TB diagnosis and elimination

Author: C Geric, Z Z Qin, C M Denkinger, S V Kik, B Marais, A Anjos, P-M David, F Ahmad Khan, A Trajman

SHARE

https://cms.qure.ai

Back

A comprehensive review comparing qXR (Qure.ai), Lunit INSIGHT CXR, and CAD4TB for TB detection using chest X-rays found that qXR outperformed or matched Lunit in multiple studies. In a triage study with 1,196 participants from Nepal and Cameroon, qXRv2 and Lunitv4.2.7 performed similarly, both achieving AUCs of 0.94, while CAD4TBv6 had a slightly lower AUC of 0.92. In a tertiary hospital study in Pakistan (2,198 participants, using sputum culture as the reference standard), qXR outperformed Lunit, with AUCs of 0.92 vs. 0.85, respectively. A meta-analysis pooling four triage studies across Pakistan, South Africa, Tanzania, and Zambia found qXR had the highest AUC (0.85) compared to Lunit (0.83) and CAD4TB (0.83). Key takeaway: While both qXR and Lunit demonstrate strong performance, qXR has shown better accuracy in certain high-burden settings, particularly in tertiary hospitals and large multi-country analyses, highlighting the importance of local calibration and validation for optimal deployment.

Authors

C Geric, Z Z Qin, C M Denkinger, S V Kik, B Marais, A Anjos, P-M David, F Ahmad Khan, A Trajman

Share this publication