Book chapter (B2)

COVID-19 detection from X-ray images using artificial intelligence

List of Authors: Subasi Abdulhamit, Qureshi Saqib Ahmed, Brahimi Tayeb, Serireti Akila

Edition name or number: 1

Publisher: Elsevier

Publication year: 2021

Book title *: Artificial Intelligence and Big Data Analytics for Smart Healthcare

Journal name in source: Artificial Intelligence and Big Data Analytics for Smart Healthcare

Title of series: Next Gen Tech Driven Personalized Med&Smart Healthcare

ISBN: 978-0-12-822060-3




Currently, the most common disease is the new coronavirus disease identified as COVID-19. Various techniques to identifying the COVID-19 disease have been offered. Computer vision techniques are widely used to classify COVID-19 with the use of chest X-ray images. Rapid clinical results may prevent COVID-19 from spreading and help doctors treat patients under high workload conditions. As the normal diagnosis phase of illness with a laboratory test is time-consuming and requires a well-equipped laboratory, the X-ray imaging technique is a fast and cheap diagnostic tool for COVID-19. Machine learning methods can enhance the diagnosis of COVID-19 as a decision support platform for radiologists. This chapter utilizes various convolutional neural network (CNN) models, including pretrained models, to classify X-ray images into three classes: COVID-19, pneumonia, and normal. CNN, a form of deep neural networks that have become dominant in various computer vision tasks, attracts interest across various domains, including radiology. Pretrained models on ImageNet are good at detecting high-level features such as edges and patterns. These models understand certain representations of features, which can be reused. Also, deep classifiers have shown promising results in many kinds of work across various domains. We drew some useful results from these classifiers, which could be used faster when detecting COVID-19. Experimental results showed that the accuracy of the VGG19 classifier is 97.56%.

Last updated on 2022-30-11 at 17:59