Automated Segmentation of Infected Regions in Chest CT Images of COVID-19 Patients using Supervised Naïve Gaussian Bayes Classifier

In this paper, one hundred chest Computed Tomography images of COVID-19 patients were used to build and test Naïve Gaussian Bayes classifier for discriminating normal from abnormal tissues. Infected areas in these images were manually segmented by an expert radiologist. Pixel grey value, local entropy and Histograms of Oriented Gradients HOG were extracted as features for tissue image classification. Based on five-folds classification experiments, the accuracy score of the classifier in this fold reached around 79.94%. Classification was more precise (85%) in recognizing normal tissue than abnormal tissue (63%). The effectiveness in identifying positive labels was also more evident with normal tissue than the abnormal one.