Coronavirus (COVID-19) is a viral disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). The spread of COVID-19 seems to have a detrimental effect on the global economy and health. A positive chest X-ray of infected patients is a crucial step in the battle against COVID-19. Early results suggest that abnormalities exist in chest X-rays of patients suggestive of COVID-19. This has led to the introduction of a variety of deep learning systems and studies have shown that the accuracy of COVID-19 patient detection through the use of chest X-rays is strongly optimistic. Deep learning networks like convolutional neural networks (CNNs) need a substantial amount of training data. Because the outbreak is recent, it is difficult to gather a significant number of radiographic images in such a short time. Therefore, in this research, we present a method to generate synthetic chest X-ray (CXR) images by developing an Auxiliary Classifier Generative Adversarial Network (ACGAN) based model called CovidGAN. In addition, we demonstrate that the synthetic images produced from CovidGAN can be utilized to enhance the performance of CNN for COVID-19 detection. Classification using CNN alone yielded 85% accuracy. By adding synthetic images produced by CovidGAN, the accuracy increased to 95%. We hope this method will speed up COVID-19 detection and lead to more robust systems of radiology.
- Deep learning,
- convolutional neural networks,
- generative adversarial networks,
- synthetic data augmentation,
- COVID-19 detection
- Gallium nitride,
- Generative adversarial networks,
- Biomedical imaging,
- X-ray imaging,
- Computer architecture,
- Machine learning
Coronavirus disease is a respiratory disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). COVID-19 was initially detected in Wuhan, China, in December 2019, and has spread worldwide since then leading to the ongoing 2020 coronavirus pandemic. More than 4.18 million cases and 286,000 deaths have been registered in more than 200 countries and territories as of 12 May 2020. Since no vaccines or cures exist, the only efficient way of human protection against COVID-19 is to reduce spread by prompt testing of the population and isolation of the infected individuals.
Certain health symptoms combined with a chest X-ray can be used to diagnose this infection. A chest X-ray can be used as a visual indicator of coronavirus infection by the radiologists. This led to the creation of numerous deep learning models, and tests have shown that it is highly likely that patients with COVID-19 infection are detected correctly by using chest radiography images.
Convolutional neural networks (CNNs) have attained state-of-the-art performance in the field of medical imaging, given enough data –. Such performance is accomplished by training on labeled data and fine-tuning its millions of parameters. CNN’s can easily overfit on small datasets because of the large number of parameters, therefore, the efficiency of generalization is proportional to the size of the labeled data. With limited quantity and variety of samples, the biggest challenge in the medical imaging domain is small datasets –. The medical image collection is a very expensive and tedious process that requires the participation of radiologists and researchers . Also, since the COVID-19 outbreak is recent, sufficient data of chest X-ray (CXR) images is difficult to gather. We propose to alleviate the drawbacks by using synthetic data augmentation. Data augmentation methods are employed to extend the training dataset artificially. Current data augmentation techniques use simple modifications to incorporate affinity like image transformations and color adjustments, such as scaling, flipping, converting, improving contrast or brightness, blurring, and sharpening, white balance, etc . This classical data augmentation is fast, reliable, and easy. However, in this augmentation, the changes are limited because it is structured to turn an existing sample into a slightly altered sample. In other words, classical data augmentation does not produce completely unseen data. A modern, advanced form of augmentation is synthetic data augmentation which overcomes the limitations of classical data augmentation. Generative Adversarial Network (GAN) is one such innovative model that generates synthetic images. It is a powerful method to generate unseen samples with a min-max game without supervision . The general concept of the GANs is to use two opposing networks (G(z) and D(x)), where one (G(z) generator) produces a realistic image to trick the other net that is equipped to better discriminate between the true and false images (D(z) discriminator). The aim of the generator is to minimize the cost value function V(D, G) whereas the discriminator maximizes it . Related works and contributions are discussed below.
In this research, we proposed an ACGAN based model called CovidGAN that generates synthetic CXR images to enlarge the dataset and to improve the performance of CNN in COVID-19 detection. The research is implemented on a dataset with 403 COVID-CXR images and 721 NormalCXR images. Our limited dataset highlights the scarcity of medical images in the research communities. Initially, the proposed CNN architecture is used to classify the two classes (that is COVID-CXR and Normal-CXR). Further, the performance of CNN with synthetic data augmentation technique is investigated. Synthetic data augmentation adds more variability to the dataset, by enlarging it. CovidGAN is used to generate synthetic images of chest X-ray (CXR). An improvement in classification performance from 85% to 95% accuracy is recorded when CNN is trained on actual data and synthetic augments. An increase in precision and recall of both the classes are also observed. Our findings show that synthesized images of CXR have significant visualizations and features that help in the detection of COVID-19. Lastly, a detailed analysis of the performance of our CNN architecture with synthetic data augmentation technique is given in Table 1. In conclusion, we proposed a way to enhance the accuracy of COVID-19 detection with minimal data by generating synthetic medical images of chest X-ray. Despite its excellent results, CovidGAN is not intended to compete with laboratory testing. Instead, we hope that this approach leads to stronger and more reliable radiology systems. In the future, we intend to improve the quality of the synthetic CXR images by training a Progressive Growing GAN .
The Kavian Scientific Research Association (KSRA) is a non-profit research organization to provide research / educational services in December 2013. The members of the community had formed a virtual group on the Viber social network. The core of the Kavian Scientific Association was formed with these members as founders. These individuals, led by Professor Siavosh Kaviani, decided to launch a scientific / research association with an emphasis on education.
KSRA research association, as a non-profit research firm, is committed to providing research services in the field of knowledge. The main beneficiaries of this association are public or private knowledge-based companies, students, researchers, researchers, professors, universities, and industrial and semi-industrial centers around the world.
Our main services Based on Education for all Spectrum people in the world. We want to make an integration between researches and educations. We believe education is the main right of Human beings. So our services should be concentrated on inclusive education.
The KSRA team partners with local under-served communities around the world to improve the access to and quality of knowledge based on education, amplify and augment learning programs where they exist, and create new opportunities for e-learning where traditional education systems are lacking or non-existent.
FULL Paper PDF file:CovidGAN: Data Augmentation Using Auxiliary Classifier GAN for Improved Covid-19 Detection
CovidGAN: Data Augmentation Using Auxiliary Classifier GAN for Improved Covid-19 Detection
in IEEE Access, vol. 8, pp. 91916-91923, 2020
PDF reference and original file: Click here
Professor Siavosh Kaviani was born in 1961 in Tehran. He had a professorship. He holds a Ph.D. in Software Engineering from the QL University of Software Development Methodology and an honorary Ph.D. from the University of Chelsea.