Predictive Model for Progression from Mild to Severe COVID-19 Using Clinical and CT Data
Overview
This multicenter study developed and externally validated a predictive model combining clinical, biological, and chest CT parameters to identify mild COVID-19 patients at risk of progression to moderate, severe, or critical disease within 30 days. The model integrates AI-based quantitative CT analysis and radiomics features to enhance risk stratification beyond traditional clinical predictors.
Background
COVID-19 patients with initially mild symptoms can rapidly deteriorate, leading to respiratory failure and high mortality, though this occurs in only about 5% of cases. Despite vaccination efforts, progression from mild to severe disease remains a concern, especially in at-risk populations. Chest CT imaging plays a key role in diagnosis and prognosis of COVID-19 pneumonia, with AI and radiomics offering potential to improve prediction of adverse outcomes. However, prior studies have focused mainly on severe cases, leaving a gap in risk stratification tools for mild COVID-19 patients presenting with respiratory symptoms.
Data Highlights
Parameter
Details
Study Population
Development cohort: 3 university hospitals + 1 private hospital in France; Validation cohort: 2 university hospitals in France
Inclusion Criteria
Adults ≥18 years, mild COVID-19 with respiratory symptoms, chest CT without contrast (development), with/without contrast (validation)
Outcome
Progression to moderate, severe, or critical COVID-19 or death within 30 days
CT Analysis
AI-based segmentation and quantification of lung opacities; radiomics features extracted and normalized
Validation
Internal validation and external validation on independent cohort with injected and non-injected CT scans
Key Findings
A predictive model combining clinical, biological, and chest CT data was successfully developed to identify mild COVID-19 patients at risk of progression.
AI-based quantitative CT analysis improved assessment of lung involvement and severity beyond visual scoring.
Radiomics features extracted from CT images contributed additional prognostic information for disease progression.
The model was externally validated in an independent cohort including patients with contrast-enhanced CT scans, demonstrating robustness.
Clinical parameters such as age, comorbidities, and symptom duration were important predictors alongside imaging features.
Clinical Implications
This predictive model can assist clinicians in early identification of mild COVID-19 patients at higher risk of clinical deterioration, enabling targeted monitoring and timely intervention. Incorporating AI-driven CT analysis and radiomics into routine assessment may optimize resource allocation and improve patient outcomes by guiding decisions on hospitalization and therapeutic strategies.
Conclusion
Integrating clinical, biological, and advanced imaging data through AI and radiomics provides a reliable tool to predict progression from mild to more severe COVID-19 forms. This approach supports personalized patient management and efficient healthcare resource use during the ongoing pandemic.
References
Original Study Authors/2024 -- Creation and external assessment of a predictive model for the progression from mild to moderate or severe COVID-19 cases
