Scholarlink

As of the end of 2020, COVID-19 has tragically resulted in over 1.8 million deaths worldwide, underscoring the urgent need for effective and accurate diagnostic methods . Chest X-ray imaging is a crucial tool in diagnosing COVID-19, but the limited availability of actual COVID-19 chest X-ray data poses a significant challenge for developing automated screening systems. Generative Adversarial Networks (GANs) and Deep Convolutional GANs (DCGANs) offer a promising solution by generating synthetic COVID-19 chest X-ray images that closely resemble real ones. In this study, we utilized GAN and DCGAN models to produce high-quality synthetic chest X-ray images, which were then used to augment the training datasets of COVID-19 classification models. Our qualitative analysis shows that these synthetic images closely resemble real X-rays, and the augmentation led to a substantial improvement in model accuracy, increasing from 94.6% to 96.8%. Our approach provides a valuable tool in combating the global pandemic by addressing data limitations and improving diagnostic accuracy, offering a pathway to more reliable and scalable COVID-19 detection.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhancing Lung Disease Detection Using DCGAN-Augmented Chest X-Ray Images for CNN Training

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Abdessamad Makan,
  • Ikram El Asri

摘要

As of the end of 2020, COVID-19 has tragically resulted in over 1.8 million deaths worldwide, underscoring the urgent need for effective and accurate diagnostic methods . Chest X-ray imaging is a crucial tool in diagnosing COVID-19, but the limited availability of actual COVID-19 chest X-ray data poses a significant challenge for developing automated screening systems. Generative Adversarial Networks (GANs) and Deep Convolutional GANs (DCGANs) offer a promising solution by generating synthetic COVID-19 chest X-ray images that closely resemble real ones. In this study, we utilized GAN and DCGAN models to produce high-quality synthetic chest X-ray images, which were then used to augment the training datasets of COVID-19 classification models. Our qualitative analysis shows that these synthetic images closely resemble real X-rays, and the augmentation led to a substantial improvement in model accuracy, increasing from 94.6% to 96.8%. Our approach provides a valuable tool in combating the global pandemic by addressing data limitations and improving diagnostic accuracy, offering a pathway to more reliable and scalable COVID-19 detection.