A Robust End-to-End CNN Architecture for Efficient COVID-19 Prediction form X-ray Images with Imbalanced Data

Abstract

The spread of coronavirus disease in late 2019 caused massive damage to human lives and forced chaos in health care systems around the globe. Early diagnosis of this disease can help separate patients from healthy people. Therefore, precise COVID-19 detection is necessary to prevent the spread of this virus thereby protecting more people form imminent death. Many advanced artificial intelligence technologies like deep learning have used chest X-ray images for this task. In this paper, we propose to classify chest X-ray images using a new end-to-end deep convolutional neural network architecture. The new model is applied on a 256 256 3 input image and consists of six convolutional blocks. Each block consists of one convolutional layer, one ReLU layer, and one max-pooling layer. Furthermore, we improve the performance of our model by adding regularization techniques, including batch normalization and dropout. The new model was applied to a challenging imbalanced COVID-19 dataset of 5000 images which consists of two classes: COVID and Non-COVID. Four metrics were used to test our new model: sensitivity, specificity, precision, and F1 score. In experiments, we achieved a sensitivity rate of 97\%, a specificity rate of 99.32\%, a precision rate of 99.90\%, and F1 score of 97.73\% despite being provided with fewer training images. In conclusion, we proposed a light deep learning model capable of achieving high prediction accuracy and outperformed state-of-the-art deep learning methods in terms of specificity and produced comparable results in terms of sensitivity.