COVID-19 diagnosis using clinical markers and multiple explainable artificial intelligence approaches: A case study...

Area of Science:

• Medical Diagnostics
• Artificial Intelligence in Medicine
• Computational Biology

Background:

• The COVID-19 pandemic necessitates accurate and rapid patient screening.
• Current diagnostic methods like RT-PCR can yield false negatives, highlighting the need for supplementary validation.
• Distinguishing COVID-19 from other respiratory illnesses based on symptoms alone is challenging.

Purpose of the Study:

• To investigate the efficacy of artificial intelligence (AI) and machine learning (ML) in diagnosing COVID-19 using clinical markers.
• To evaluate the performance of various AI classifiers in identifying COVID-19.
• To assess the interpretability of AI predictions using explainable AI techniques.

Main Methods:

• Utilized clinical markers (eosinophils, lymphocytes, red blood cells, leukocytes) as input features for AI models.
• Trained and evaluated multiple machine learning classifiers for COVID-19 diagnosis.
• Applied five explainable AI techniques to interpret model predictions.
• The k-nearest neighbor algorithm was among the classifiers tested.

Main Results:

• The k-nearest neighbor algorithm achieved the highest performance metrics: 84% accuracy, 85% precision, 84% recall, and 84% F1-score.
• Specific clinical markers, including eosinophils, lymphocytes, red blood cells, and leukocytes, were identified as significant in differentiating COVID-19.
• Explainable AI techniques provided insights into the decision-making processes of the diagnostic models.

Conclusions:

• AI and ML models, particularly k-nearest neighbor, show significant potential for improving COVID-19 diagnostic accuracy.
• The combination of specific clinical markers offers a robust approach for distinguishing COVID-19.
• Integrating AI-based classifiers with RT-PCR testing can lead to more reliable and efficient disease diagnosis.