Equilibrium-based COVID-19 diagnosis from routine blood tests: A sparse deep convolutional model.

Area of Science:

• Medical Informatics
• Biomedical Engineering
• Computational Biology

Background:

• The COVID-19 pandemic necessitates rapid and reliable diagnostic tools for clinical decision-making.
• Predictive models can alleviate healthcare system strain through early and accurate patient screening.
• Routine blood tests offer potential advantages for initial COVID-19 screening.

Purpose of the Study:

• To propose a novel COVID-19 prediction model utilizing routine blood tests.
• To develop an efficient feature selection mechanism for blood test data.
• To implement a 1D Convolutional Neural Network (1DCNN) for COVID-19 diagnosis.

Main Methods:

• A sparsification procedure to exploit feature dependencies in blood test data.
• A hybrid feature selection mechanism combining Pearson correlation and a novel Minkowski-based Equilibrium Optimizer (MEO).
• A 1D Convolutional Neural Network (1DCNN) for final diagnostic classification.

Main Results:

• The proposed model achieved an average testing accuracy of 98.5% on the OSR dataset.
• The model effectively diagnosed COVID-19 using less than half of the available blood tests.
• Outperformed existing state-of-the-art techniques in COVID-19 prediction accuracy.

Conclusions:

• The developed model demonstrates high accuracy and efficiency in COVID-19 prediction using routine blood tests.
• This approach offers a promising tool for early screening and clinical decision support.
• Reducing the number of required blood tests can streamline the diagnostic process.