Added value of chest CT in a machine learning-based prediction model to rule out COVID-19 before inpatient admission: A retrospective university network study

Area of Science:

• Medical Imaging and Artificial Intelligence
• Infectious Disease Diagnostics
• Machine Learning in Healthcare

Background:

• Timely identification of COVID-19 in patients awaiting hospital admission remains a challenge.
• Artificial intelligence (AI) offers potential solutions for improving pandemic preparedness.
• Leveraging established clinical networks can facilitate data collection and model development.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting COVID-19 infection before inpatient admission.
• To assess the added value of chest CT features in AI-based diagnostic models.
• To enhance prospective pandemic preparedness through improved early identification of infected individuals.

Main Methods:

• Supervised machine learning was employed to construct diagnostic models.
• A retrospective pooled database of 4437 participants from 12 German centers was utilized.
• Models were trained using chest CT features (R), clinical and laboratory data (CL), or a combination (RCL).

Main Results:

• Models integrating chest CT features demonstrated significantly improved performance.
• The combined model (RCL) achieved higher sensitivity (0.89) and specificity (0.89) compared to clinical data alone (CL: sensitivity 0.82, specificity 0.84).
• The area under the receiver operating curve (AUC) for model RCL was 0.95, indicating superior diagnostic accuracy.

Conclusions:

• Incorporating chest CT features into machine learning models substantially enhances the ability to rule out COVID-19 prior to hospital admission.
• AI-driven diagnostic tools, particularly those utilizing imaging data, can improve the efficiency of patient triage.
• This approach holds promise for strengthening hospital readiness during infectious disease outbreaks.