COVID-19 detection using machine learning and fusion-based deep learning models

Abstract

The COVID-19 pandemic has been one of the most challenging crises attacking the world in the last three years. Many systems have been introduced in the field of COVID-19 detection. In this research, machine learning and deep learning models for the detection of COVID-19 with a probability of the presence of COVID-19 are proposed. In the machine learning scenario, the COVID-19 dataset is split into 70% training and 30% testing, and a segmentation process is applied to the CT images in order to get the lung ROI only. The features of CT images are then extracted using Gabor-Wavelet and deep-based features. The SVM classifier is then trained and evaluated. For the deep learning model, the CT images are fed into the model without feature extraction, and three different DL models (CNN, GoogleNet, and ResNet50) are trained and evaluated. Other scenarios are proposed in which the SVM Gabor-Wavelet and deep features are fused, and the three deep learning models are also fused to get better performance. The experiments show that the best model is the deep-based fusion model by which the system achieved 96.4156%, 96.1905%, and 96.1905% for accuracy, precision, and recall, respectively.