Performance of AI Approaches for COVID-19 Diagnosis Using Chest CT Scans: The Impact of Architecture and Dataset
RöFo - Fortschritte auf dem Gebiet der Röntgenstrahlen und der bildgebenden Verfahren · 01.02.2026 · 198 (02) · 185-198
DOI: 10.1055/a-2577-3928 ↗Kurz gefasst
Multizentrische Studie in der Zeitschrift RöFo - Fortschritte auf dem Gebiet der Röntgenstrahlen und der bildgebenden Verfahren, erschienen am 01.02.2026. Verfasst von Astha Jaiswal und 11 weiteren Autor:innen. Beteiligte Einrichtung: Institute for Diagnostic and Interventional Radiology, Faculty of Medicine and University Hospital Cologne, University of Cologne, Cologne, Germany. Hauptthemen laut MEDLINE: COVID-19, Tomography, X-Ray Computed, Artificial Intelligence, Radiography, Thoracic. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 44 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Radiology and Technology Innovation Center of Jilin Province, NIEHS NIH HHS, RACOON (NUM), "NUM 2.0".
Abstract (Originalsprache)
AI is emerging as a promising tool for diagnosing COVID-19 based on chest CT scans. The aim of this study was the comparison of AI models for COVID-19 diagnosis. Therefore, we: (1) trained three distinct AI models for classifying COVID-19 and non-COVID-19 pneumonia (nCP) using a large, clinically relevant CT dataset, (2) evaluated the models’ performance using an independent test set, and (3) compared the models both algorithmically and experimentally. In this multicenter multi-vendor study, we collected n=1591 chest CT scans of COVID-19 (n=762) and nCP (n=829) patients from China and Germany. In Germany, the data was collected from three RACOON sites. We trained and validated three COVID-19 AI models with different architectures: COVNet based on 2D-CNN, DeCoVnet based on 3D-CNN, and AD3D-MIL based on 3D-CNN with attention module. 991 CT scans were used for training the AI models using 5-fold cross-validation. 600 CT scans from 6 different centers were used for independent testing. The models’ performance was evaluated using accuracy (Acc), sensitivity (Se), and specificity (Sp). The average validation accuracy of the COVNet, DeCoVnet, and AD3D-MIL models over the 5 folds was 80.9%, 82.0%, and 84.3%, respectively. On the independent test set with n=600 CT scans, COVNet yielded Acc=76.6%, Se=67.8%, Sp=85.7%; DeCoVnet provided Acc=75.1%, Se=61.2%, Sp=89.7%; and AD3D-MIL achieved Acc=73.9%, Se=57.7%, Sp=90.8%. The classification performance of the evaluated AI models is highly dependent on the training data rather than the architecture itself. Our results demonstrate a high specificity and moderate sensitivity. The AI classification models should not be used unsupervised but could potentially assist radiologists in COVID-19 and nCP identification.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- Institute for Diagnostic and Interventional Radiology, Faculty of Medicine and University Hospital Cologne, University of Cologne, Cologne, Germany.
- Department of Radiology, The First Hospital of Jilin University, Changchun, China.
- Department of Diagnostic and Interventional Radiology, University Hospital Heidelberg, University of Heidelberg, Heidelberg, Germany.
- Institute for Diagnostic and Interventional Radiology, Frankfurt University Hospital, Frankfurt, Germany.
- Philips Healthcare, Innovative Technologies, Aachen, Germany.
ORCID-Profile: Jonathan Kottlors, Thorsten Persigehl
Förderung
- Radiology and Technology Innovation Center of Jilin Province
- NIEHS NIH HHS
- RACOON (NUM), "NUM 2.0"
- Sino-German Center for Research Promotion (SGC), a project entitled CT-based Deep Learning Algorithm in Diagnosis and evaluation of COVID-19: An International…
- Jilin Provincial Key Laboratory of Medical imaging & big data
- China International Medical Foundation, Imaging Research, SKY
- RACOON (NUM), „NUM 2.0“
Zitieren
APA
Jaiswal, A., Fervers, P., Meng, F., Zhang, H., Móré, D., Giannakis, A., Wailzer, J., Bucher, A. M., Maintz, D., Kottlors, J., Shahzad, R., & Persigehl, T. (2026). Performance of AI Approaches for COVID-19 Diagnosis Using Chest CT Scans: The Impact of Architecture and Dataset. RöFo - Fortschritte auf dem Gebiet der Röntgenstrahlen und der bildgebenden Verfahren, 198(02), 185-198. https://doi.org/10.1055/a-2577-3928
Vancouver
Jaiswal A, Fervers P, Meng F, Zhang H, Móré D, Giannakis A, et al. Performance of AI Approaches for COVID-19 Diagnosis Using Chest CT Scans: The Impact of Architecture and Dataset. RöFo - Fortschritte auf dem Gebiet der Röntgenstrahlen und der bildgebenden Verfahren. 2026;198(02):185-198. doi:10.1055/a-2577-3928
BibTeX
@article{Jaiswal2026,
author = {Jaiswal, Astha and Fervers, Philipp and Meng, Fanyang and Zhang, Huimao and Móré, Dorottya and Giannakis, Athanasios and Wailzer, Jasmin and Bucher, Andreas Michael and Maintz, David and Kottlors, Jonathan and Shahzad, Rahil and Persigehl, Thorsten},
title = {Performance of AI Approaches for COVID-19 Diagnosis Using Chest CT Scans: The Impact of Architecture and Dataset},
journal = {RöFo - Fortschritte auf dem Gebiet der Röntgenstrahlen und der bildgebenden Verfahren},
year = {2026},
volume = {198},
number = {02},
pages = {185--198},
doi = {10.1055/a-2577-3928}
}
EUVIMED