A new method for chest X-ray images categorization using transfer learning and CovidNet_2020 employing convolution neural network

• Published in: Soft computing (Berlin, Germany) Germany), 2023-10, Vol.27 (19), p.14241-14251
• Main Authors: Raghavendran, P. S., Ragul, S., Asokan, R., Loganathan, Ashok Kumar, Muthusamy, Suresh, Mishra, Om Prava, Ramamoorthi, Ponarun, Sundararajan, Suma Christal Mary
• Format: Article
• Language: English
• Subjects: Accuracy; Alzheimer's disease; Application of Soft Computing; Artificial Intelligence; Artificial neural networks; Chest; Classification; Computational Intelligence; Computer vision; Control; COVID-19; Datasets; Deep learning; Diagnosis; Engineering; Human error; Learning; Lungs; Mathematical Logic and Foundations; Mechatronics; Medical imaging; Neural networks; Pneumonia; Robotics; Viral diseases; Viruses; X-rays
• ISSN: 1432-7643; 1433-7479
• DOI: 10.1007/s00500-023-08874-7

Since computer vision has been a very emerging and happening approach in image categorization, this article describes how chest X-ray images of diverse infected and normal samples were classified using convolution neural networks, mostly under the following five categories: normal or no lung infection, COVID-19, SARS, ARDS and other pneumonia infections such as viral pneumonia, cavitating pneumonia, streptococcus pneumonia, legionella pneumonia and pneumocystis pneumonia. The proposed approach accepts the X-ray image inputs and diagnoses the lung infection under the aforementioned five categories of major pneumonia infections. Pre-trained models like VGG19, Resnet-50, and NasNetMobile are applied to the diagnosis of chest X-ray images, and their performance is compared against a newly proposed convolutional neural network called the COVNET 2020, since the network is inspired to identify the COVID-19 chest X-ray images. Transfer learning with pre-rained neural networks on massive image databases like “imagenet” has become de facto for medical image diagnosis.