Screening of COVID-19 Based on GLCM Features from CT Images Using Machine Learning Classifiers

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Radiology

Background:

• Accurate and rapid diagnosis of COVID-19 is essential for effective patient management and public health.
• Manual interpretation of chest CT scans for COVID-19 can be challenging due to image variations and dataset limitations.
• Radiomics, a quantitative image analysis technique, shows promise for disease diagnosis and prognosis.

Purpose of the Study:

• To evaluate the effectiveness of a machine learning (ML) model utilizing Gray-Level Co-occurrence Matrix (GLCM) features from chest CT images for COVID-19 screening.
• To compare the performance of different ML classifiers, including Support Vector Machines (SVM), K-nearest neighbors (KNN), Random Forest (RF), and XGBoost, for COVID-19 detection.
• To assess the impact of hyperparameter tuning and cross-validation on the diagnostic accuracy of the ML models.

Main Methods:

• Extraction of GLCM radiomics features from low-resolution chest CT images.
• Implementation and training of multiple ML classifiers: SVM, KNN, Random Forest, and XGBoost.
• Hyperparameter tuning using validation tests and performance evaluation with tenfold cross-validation.

Main Results:

• The ML models demonstrated high performance in classifying COVID-19 cases based on CT image features.
• Random Forest and SVM classifiers achieved the highest diagnostic accuracy, reaching 99.94%.
• The study assessed model performance using key metrics including sensitivity, accuracy, and specificity.

Conclusions:

• Machine learning models employing GLCM features from chest CT scans are highly effective for COVID-19 screening.
• Random Forest and SVM algorithms show superior performance, offering a reliable and accurate computer-assisted diagnosis method.
• This approach provides a promising tool for rapid and accurate COVID-19 detection, addressing challenges in manual diagnosis.