Computed Tomography Image Analysis on COVID-19 Cases using Machine Learning Approaches

Abstract

Coronavirus disease (COVID-19) has become a serious worldwide health concern affecting the respiratory system since December 2019. Computed Tomography (CT) image analysis and identification are powerful tools for diagnosing COVID-19. However, due to the disparity of the distribution and form of COVID-19 infection and the diverse degrees of infection severity, the classification of the CT images is challenging, especially manually. Therefore, this study aims to employ artificial intelligence techniques, including machine learning and deep learning algorithms, to classify COVID-19 from CT images. The grey-level co-occurrence matrix (GLCM) features were computed and fed into machine learning classifiers, namely Support Vector Machine, Random Forest, K-Nearest Neighbour, Logistic Regression, and Naïve Bayes model for training purposes. Deep learning models, including ResNet50, Densenet121, Inception, and VGG16, were trained using raw data scaled and transformed to greyscale mode. The performances of machine learning and deep learning models were assessed on the testing data. Random Forests, ResNet50, and DenseNet121 outperform all the other models by achieving 100% accuracy, precision, sensitivity, and specificity on the dataset applied. The performance of machine learning models can be further improved by obtaining the optimised parameters in future research.