Scholarlink

The urgency of rapid and reliable diagnostic tools in the circumstances of the COVID-19 coronavirus pandemic has gained a greater emphasis for regions with limited access to laboratory tests. Chest radiography has been considered a useful support mode in the detection of COVID-19 because it is readily available and quick, unlike RT-PCR tests, but determining on X-rays whether a person has COVID-19 or other respiratory problems is difficult and often requires a lot of experience and time before making decisions that can shift to clinical importance. This study aims to address this challenge by developing an accurate model to detect COVID-19 using chest radiograph images, thus reducing the dependency on the radiologist’s expertise and speeding diagnosis. A hybrid model that combines the pre-trained VGG-16 and VGG-19 with cascaded atrous convolution (Hybrid VGG-AtrousNet) was used to extract the features for better classification. The proposed model is compared with traditional VGG-16 and VGG-19-based feature extraction, where the proposed model achieves 99% accuracy. While the models shows promising results, the future work will focus on testing its robustness across real-world datsets thereby addressing the challenges of interpretability and generalizability.

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

A Hybrid Deep Learning Approach Using Cascaded Atrous Convolution and VGG Models for COVID-19 Diagnosis

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • S. P. Mrinal Ashwith,
  • Rajalakshmi Shenbaga Moorthy

摘要

The urgency of rapid and reliable diagnostic tools in the circumstances of the COVID-19 coronavirus pandemic has gained a greater emphasis for regions with limited access to laboratory tests. Chest radiography has been considered a useful support mode in the detection of COVID-19 because it is readily available and quick, unlike RT-PCR tests, but determining on X-rays whether a person has COVID-19 or other respiratory problems is difficult and often requires a lot of experience and time before making decisions that can shift to clinical importance. This study aims to address this challenge by developing an accurate model to detect COVID-19 using chest radiograph images, thus reducing the dependency on the radiologist’s expertise and speeding diagnosis. A hybrid model that combines the pre-trained VGG-16 and VGG-19 with cascaded atrous convolution (Hybrid VGG-AtrousNet) was used to extract the features for better classification. The proposed model is compared with traditional VGG-16 and VGG-19-based feature extraction, where the proposed model achieves 99% accuracy. While the models shows promising results, the future work will focus on testing its robustness across real-world datsets thereby addressing the challenges of interpretability and generalizability.