Clinical and Laboratory Approach to Diagnose COVID-19 Using Machine Learning.

Area of Science:

• Medical Diagnostics
• Artificial Intelligence in Healthcare
• Infectious Disease Research

Background:

• Coronavirus 2 (SARS-CoV-2), causing COVID-19, significantly impacts global health and economy.
• Conventional RT-PCR testing for COVID-19 can yield false negatives and errors.
• Emerging diagnostic methods include imaging, blood tests, and cough sound analysis.

Purpose of the Study:

• To predict COVID-19 diagnosis using blood tests and machine learning.
• To review existing machine learning applications for COVID-19 diagnosis from clinical markers.
• To identify critical blood parameters indicative of COVID-19 infection.

Main Methods:

• Utilized four distinct classifiers and Synthetic Minority Oversampling Technique (SMOTE) for classification.
• Employed Shapley Additive Explanations (SHAP) to determine feature importance.
• Analyzed blood test results and clinical laboratory markers.

Main Results:

• Eosinophils, monocytes, leukocytes, and platelets were identified as the most critical blood parameters for COVID-19 diagnosis in the dataset.
• The developed classifiers demonstrated potential for improving diagnostic sensitivity.
• Machine learning models showed promise in identifying COVID-19 infection from blood markers.

Conclusions:

• Machine learning models utilizing blood parameters can aid in COVID-19 diagnosis.
• This automated framework can assist clinicians in patient screening and diagnosis.
• The approach offers a valuable adjunct to RT-PCR, particularly in emergency situations and potential future outbreaks.