A detailed and comprehensive performance analysis on the largest covid-19 chest x-ray image dataset using a custom cnn model and state-of-the-art deep learning models
Loading...
Date
2022
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
UMT Lahore
Abstract
The modern Scientific World has ever been endeavoring towards battling and devising solutions for the newly arising pandemics. One such pandemic which has disarrayed the world’s accustomed routine upside down is COVID-19 and has devastated the world’s economy consuming around 45 million lives around the globe. The governments and scientists have taken front lines striving towards the diagnosis and engineering of the vaccination for the said virus. The COVID-19 can be diagnosed using Artificial Intelligence with an accuracy higher than the traditional methods using chest X-Rays. The subject research work involves an evaluation of the performance of Deep Learning models for COVID-19 diagnosis using chest X-Ray images on a dataset containing the largest number of COVID-19 images ever used in the literature research work according to the best of the authors’ knowledge. Further, a CNN model was developed, named Custom-Model in this study for an evaluation and comparison with the state-of-the-art deep learning models. The intention was not to develop a new highperforming deep learning model rather evaluate the performance of deep learning models on a larger COVID-19 chest X-Ray images’ dataset. Moreover, Xception and MobilNetV2 based models have also been used for evaluation purposes. The criteria for evaluation have been based on Accuracy, Precision, Recall, F1-Score, ROC curves, AUC, Confusion matrix, and Macro and Weighted averages. Among the deployed models, Xception turns out to be a top performer in terms of precision and accuracy while MobileNetV2 based model could detect slightly more COVID-19 cases than Xception and slightly fewer False Negatives while giving far more False Positives than the remaining models. Also, the custom CNN model exceeds MobileNetV2 model in terms of precision. The best Accuracy, Precision, Recall, and F1-Score out of these three models turn out to be 94.2%, 99%, 95%, and 97% as shown by the Xception model. Finally, It was found that the overall accuracy in the current evaluation was curtailed by approximately 2.2% as compared to the average accuracy of previous works on multi-class
vii
classification while a very high precision value has been observed which is of high scientific value.