Detection of COVID-19 Using EfficientNet-B3 CNN and Chest Computed Tomography Images

COVID-19 has spread throughout the world. At the beginning of 2020, the World Health Organization (WHO) declared a global emergency. With this epidemic, health systems in many countries of the world are facing a major challenge. Early diagnosis and fast isolation of COVID-19 patients are critical and have proven to be effective in limiting the disease and reducing its rapid spread. Although the RT-PCR laboratory test is considered the basic diagnostic tool for the disease, the possibility of error and delayed results have made Computed Tomography (CT) a valuable diagnostic tool. Regarding this perspective, we propose a solution to diagnose patients with COVID-19 infection from CT images and machine learning. Our system is based on a new family of CNN models called EfficientNet. In particular, we modify the Efficientnet-B3 model by removing the top layer and adding our own two branches with different layers. The proposed method is tested on the SARS-CoV-2 CT dataset and has achieved impressive results compared to recent state-of-the-art solutions, with its overall accuracy reaching 99% when using 80% of the data for training.