DenTcov: Deep Transfer Learning-Based Automatic Detection of Coronavirus Disease (COVID-19) Using Chest X-ray Images

El Idrissi El-Bouzaidi, Y.; Abdoun, O.

On 31 December 2019, COVID-19, a novel coronavirus, appeared for the first time in the Chinese city of Wuhan, to act as a preliminary warning and affected a wider human being in the world. This virus, declared a pandemic by the auspices of the World Health Organization (WHO), given its high rate of transmissibility. The protocol most often used to detect the virus is PCR. It is a time-consuming and less sensitive procedure with high false-negative results. These problems are solved through radiographic imaging techniques to detect radioactive symptoms related to COVID-19. Furthermore, significant time is required to complete the analytical task, and mistakes can occur, meaning that automation is necessary. The use of advanced Artificial intelligence tools can significantly accelerate both the time and quality of the analysis. We suggest DenTcov, a computer-aided approach to detect COVID-19 infection via chest X-ray images. Our model is a two-phase process: Phase (1) Pre-Processing and data augmentation, Phase (2) COVID-19 detection based DensNet121, a pre-trained model, then trained with the dataset prepared by us. During the experimental phase of DenTcov, we measure the performances of the architecture by calculating a set of common metrics, both 2-class and 3-class classification. The experimental assessment confirms the DenTcov model offers a 96.52 and 99% higher classification accuracy for three and two classes, respectively, compared to other proposed methodologies. © 2021, The Author(s), under exclusive license to Springer Nature Switzerland AG.