A Novel Medical Diagnosis model for COVID-19 infection detection based on Deep Features and Bayesian Optimization

Area of Science:

• * Artificial Intelligence
• * Medical Imaging Analysis
• * Infectious Disease Diagnostics

Background:

• * Coronavirus disease 2019 (COVID-19) emerged as a global pandemic, causing significant public health challenges.
• * Accurate and rapid detection of COVID-19 is crucial for effective clinical management and control.
• * Existing diagnostic methods may require further support for widespread application.

Purpose of the Study:

• * To develop an intelligent computer-aided model for automatic detection of COVID-19 positive cases.
• * To leverage deep Convolutional Neural Networks (CNNs) for feature extraction from chest X-ray images.
• * To integrate extracted features with machine learning algorithms for enhanced diagnostic accuracy.

Main Methods:

• * A novel five-layer deep CNN architecture was designed and trained from scratch for feature extraction.
• * Extracted deep features were utilized with k-nearest neighbor, Support Vector Machine (SVM), and decision tree classifiers.
• * Hyperparameters for machine learning models were optimized using the Bayesian optimization algorithm on a public COVID-19 radiology database.

Main Results:

• * The Support Vector Machine (SVM) classifier achieved the highest performance.
• * Achieved accuracy of 98.97%, sensitivity of 89.39%, specificity of 99.75%, and an F-score of 96.72%.
• * The developed model demonstrated high reliability in identifying COVID-19 cases from chest X-rays.

Conclusions:

• * A cost-effective, rapid, and reliable AI tool for COVID-19 detection has been developed.
• * The model can assist healthcare professionals, including physicians and radiologists, in decision-making.
• * The tool has the potential to reduce misdiagnosis rates and serve as a retrospective validation tool.