[Paper Review] Detection of Coronavirus (COVID-19) Associated Pneumonia based on Generative Adversarial Networks and a Fine-Tuned Deep Transfer Learning Model using Chest X-ray Dataset

Nour Eldeen M. Khalifa, Mohamed Hamed N. Taha+2 more|arXiv (Cornell University)|Apr 2, 2020

COVID-19 diagnosis using AIMedicine61 citations

TL;DR

The paper uses GAN-based data augmentation and fine-tuned deep transfer learning models to detect COVID-19-associated pneumonia from chest X-rays, with ResNet18 achieving 99% testing accuracy on a limited dataset.

ABSTRACT

The COVID-19 coronavirus is one of the devastating viruses according to the world health organization. This novel virus leads to pneumonia, which is an infection that inflames the lungs' air sacs of a human. One of the methods to detect those inflames is by using x-rays for the chest. In this paper, a pneumonia chest x-ray detection based on generative adversarial networks (GAN) with a fine-tuned deep transfer learning for a limited dataset will be presented. The use of GAN positively affects the proposed model robustness and made it immune to the overfitting problem and helps in generating more images from the dataset. The dataset used in this research consists of 5863 X-ray images with two categories: Normal and Pneumonia. This research uses only 10% of the dataset for training data and generates 90% of images using GAN to prove the efficiency of the proposed model. Through the paper, AlexNet, GoogLeNet, Squeeznet, and Resnet18 are selected as deep transfer learning models to detect the pneumonia from chest x-rays. Those models are selected based on their small number of layers on their architectures, which will reflect in reducing the complexity of the models and the consumed memory and time. Using a combination of GAN and deep transfer models proved it is efficiency according to testing accuracy measurement. The research concludes that the Resnet18 is the most appropriate deep transfer model according to testing accuracy measurement and achieved 99% with the other performance metrics such as precision, recall, and F1 score while using GAN as an image augmenter. Finally, a comparison result was carried out at the end of the research with related work which used the same dataset except that this research used only 10% of original dataset. The presented work achieved a superior result than the related work in terms of testing accuracy.

Motivation & Objective

• Motivate detection of COVID-19 pneumonia from chest X-rays.
• Address limited data by using GAN-based image generation to augment the dataset.
• Evaluate lightweight transfer learning models for pneumonia detection.
• Identify which model among AlexNet, GoogLeNet, SqueezeNet, and ResNet18 performs best.
• Demonstrate robustness against overfitting through GAN augmentation.

Proposed method

• Use Generative Adversarial Networks to generate additional chest X-ray images to augment training.
• Apply fine-tuned deep transfer learning on a dataset of 5863 X-ray images categorized as Normal or Pneumonia.
• Evaluate four lightweight architectures (AlexNet, GoogLeNet, SqueezeNet, ResNet18) for pneumonia detection.
• Train with 10% of the dataset and generate 90% synthetic images via GAN to test robustness and performance.
• Measure performance using accuracy and additional metrics (precision, recall, F1).
• Determine the best-performing model based on testing accuracy.

Experimental results

Research questions

• RQ1 Can GAN-generated images effectively augment a small chest X-ray dataset for pneumonia detection?
• RQ2 Which lightweight deep transfer learning model yields the highest testing accuracy for COVID-19 pneumonia detection on this augmented dataset?
• RQ3 Does GAN augmentation reduce overfitting and improve performance across the evaluated models?
• RQ4 What are the comparative performance metrics (precision, recall, F1) for Normal vs Pneumonia classes with GAN augmentation?

Key findings

• GAN augmentation with a small original dataset can achieve high testing accuracy.
• Among the models, ResNet18 delivered the best performance.
• The approach achieved 99% testing accuracy with associated precision, recall, and F1-score on the augmented dataset.
• The study reports superiority over related work using the same dataset under 10% training data.
• GAN-based augmentation improves model robustness against overfitting in this setting.