Related Experiment Video
Updated: May 16, 2025
08:05
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
14.0K
Development and multicentric external validation of a prognostic COVID-19 severity model based on thoracic CT
Ine Dirks1,2, Matías Nicolás Bossa3, Abel Díaz Berenguer3
1Department of Electronics and Informatics (ETRO), Vrije Universiteit Brussel (VUB), Pleinlaan, Brussels, 1050, Belgium. ine.dirks@vub.be.
BMC Medical Informatics and Decision Making
|April 1, 2025
Summary
A new model using thoracic computed tomography (CT) features and patient data can predict severe COVID-19. This tool offers rapid risk stratification to aid clinical decisions and resource allocation during high-incidence periods.
Area of Science:
- Radiology
- Medical Informatics
- Computational Biology
Background:
- Effective risk stratification of COVID-19 patients is crucial for timely therapeutic decisions, hospital resource allocation, and patient management.
- A single-source data model can provide efficient decision support during periods of high disease incidence.
Purpose of the Study:
- To develop and validate a prognostic model for identifying COVID-19 patients at risk of severe disease within one month.
- To assess the model's performance across different stages of the pandemic, including the prevalence of various SARS-CoV-2 variants.
Main Methods:
- A logistic regression model was developed using patient age, sex, and imaging features extracted from thoracic computed tomography (CT) scans.
- The model was trained on the Study of Thoracic CT in COVID-19 (STOIC) challenge dataset and validated on independent internal and external multicentric datasets.
- Performance was evaluated separately on data acquired before and during the dominance of delta and omicron variants.
Main Results:
- Logistic regression using handcrafted features performed comparably to deep learning approaches, offering a simpler solution.
- Predictive features included lesion and lung parenchyma characteristics, alongside patient age and sex.
- The model achieved an area under the curve (AUC) of 0.78 on the challenge test set and 0.74 on the external test set, with stable performance across different pandemic periods.
Conclusions:
- Thoracic CT features and metadata can be utilized in a logistic regression model for rapid estimation of short-term COVID-19 severity, enabling decision support.
- The model's consistent performance indicates its potential for seamless clinical integration, particularly for early decision-making and resource optimization during COVID-19 surges.
- This study establishes a foundation for prognostic modeling in respiratory infections using readily available imaging data.
