COV19X-Net: A Convolutional Neural Network-based method for classifying COVID-19 in chest X-ray images

The SARS-CoV-2 virus is the infectious pathogen responsible for COVID-19. Early COVID-19 diagnosis is crucial for the proper treatment of the infected patient. Popular detection method like-RT-PCR can significantly reduce the effectiveness of diagnosis and it is costly and insensitive to diagnose COVID-19. That is why an automated approach for the detection of COVID-19 from medical images is essential to make accurate, fast and cost-effective prediction. This study introduces the application of a proposed model named COV19X-Net, based on InceptionResNetV2 a pre-trained Convolutional Neural Network architecture. COV19X-Net is trained and tested on three chest X-ray datasets. COVID-19 is one of the four classes of chest radiography pictures in these datasets. To our knowledge, there have been a very few comparisons of automated approaches for 4-class COVID-19 classification on diverse datasets. As a result, this study has evaluated the efficacy of the proposed model COV19X-Net to that of previously well known research works using a variety of metrics including accuracy, precision, recall rate and F1 score. With accuracy of 86.36%, 94.42% and 91.88% respectively on three datasets, the suggested model COV19X-Net has performed better on test data that has not been pre-processed.