Multi-Modal Machine Learning Approach for COVID-19 Detection Using Biomarkers and X-Ray Imaging

Area of Science:

• Artificial Intelligence in Medicine
• Machine Learning for Diagnostics
• Medical Imaging Analysis

Background:

• Accurate COVID-19 detection is crucial, particularly in resource-limited areas.
• Existing diagnostic methods have limitations in speed and reliability.
• AI models integrating diverse data sources are needed to complement current diagnostics.

Purpose of the Study:

• To develop and evaluate a multi-modal machine learning model for COVID-19 diagnosis.
• To combine clinical biomarkers and chest X-ray images for enhanced accuracy.
• To provide interpretable insights into the diagnostic process.

Main Methods:

• A dataset of 250 patients (180 COVID-19 positive, 70 negative) was utilized.
• Clinical biomarkers (CRP, ferritin, NLR, albumin) and chest X-rays were analyzed.
• A late-fusion strategy integrated Gradient Boosting (biomarkers) and VGG CNN (images) models.

Main Results:

• The Gradient Boosting + VGG fusion model achieved high performance: AUC-ROC 0.94, F1-score 0.93.
• Key performance metrics included Specificity of 93% and NPV of 96%.
• Interpretability analysis highlighted CRP, ferritin, and lung regions as significant predictors.

Conclusions:

• The multi-modal AI model significantly improves COVID-19 diagnostic accuracy over single-modality approaches.
• Interpretability of the model aligns with clinical knowledge, supporting its practical application.
• This approach offers a valuable tool for rapid and reliable COVID-19 detection in clinical practice.