SARS-CoV-2 Prediction Strategy Based on Classification Algorithms from a Full Blood Examination.

Area of Science:

• Medical Diagnostics
• Computational Biology
• Infectious Disease Research

Background:

• Accurate and rapid diagnosis of infectious diseases like SARS-CoV-2 is crucial for controlling spread.
• Current methods like RT-PCR and RDT have limitations.
• Distinguishing COVID-19 from other infections requires improved diagnostic strategies.

Purpose of the Study:

• To train and optimize Support Vector Machine (SVM) and K-nearest neighbors (KNN) classifiers.
• To enable rapid identification of SARS-CoV-2 positive/negative patients using complete blood tests.
• To assess diagnostic performance without prior patient health information.

Main Methods:

• Utilized SVM and KNN machine learning models.
• Applied models to patient data from Israelita Albert Einstein, São Paulo.
• Analyzed two patient groups: 'regular ward' and 'not admitted to the hospital'.
• Selected blood profiles using the ANOVA statistical test of dependence.

Main Results:

• Both SVM and KNN models provided early and accurate detection of SARS-CoV-2.
• The optimized SVM technique demonstrated superior performance in non-hospitalized patients.
• Achieved high performance metrics: 94% precision, 96% recall, 95% accuracy, and 99% AUC for SVM in non-hospitalized patients.

Conclusions:

• Machine learning models trained on blood profiles offer a promising strategy for rapid infectious disease screening.
• This approach can significantly improve initial screening for SARS-CoV-2.
• The study highlights the potential of complete blood tests combined with AI for public health.