Related Experiment Video
Updated: Jul 17, 2026
08:05
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
14.1K
Multi-centre benchmarking of deep learning models for COVID-19 detection in chest x-rays
Rachael Harkness1,2, Alejandro F Frangi3,4, Kieran Zucker5
1School of Computing, University of Leeds, Leeds, United Kingdom.
Frontiers in Radiology
|June 5, 2024
Summary
Deep learning models for COVID-19 detection from chest X-rays performed adequately on UK populations but poorly internationally. Clinician involvement is crucial for developing clinically useful AI tools.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiology
Background:
- Deep learning models show promise for COVID-19 detection using chest X-rays.
- Evaluating these models' clinical utility as decision support tools is essential.
Purpose of the Study:
- To retrospectively evaluate deep learning models for COVID-19 detection from chest X-rays.
- To assess the suitability of these AI systems as clinical decision support tools.
Main Methods:
- Models trained on the National COVID-19 Chest Imaging Database (NCCID) and evaluated on independent international datasets.
- Analysis included clinical/technical error contributors, model bias, and explainable prediction techniques.
Main Results:
- Models performed comparably to radiologists on UK (NHS) populations but generalized poorly internationally.
- Performance varied by sex and age, and models failed on complex cases and mild COVID-19 presentations.
Conclusions:
- Current AI model development practices have significant pitfalls, leading to impractical tools.
- Clinician involvement throughout the AI development lifecycle is imperative for creating fit-for-purpose diagnostic systems.
