Scholarlink

We develope a pneumoconiosis classification model for chest X-ray images using Vision Transformer(ViT) and compare its accuracy with that of Convolutional Neural Networks (CNNs). Since there is variability in accuracy and differences in the accuracy across models, three ViT pretrained models are used for verification. Common CNN models, VGG16 and ResNet50, are used as comparison models. Then trials are performed for each training condition, and the average correct response rate is used as the evaluation index. The training results have shown that ViT is as accurate as or more accurate than CNN in all training conditions. This suggests that ViT is useful for pneumoconiosis classification. In terms of training time, ViT requires less time than CNN for a small number of epochs.

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

Vision Transformer Model for Lung Disease Classification Using Chest X-Ray Images

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ryota Ueki,
  • Yoshua Kazukuni Nomura,
  • Narufumi Suganuma,
  • Shinichi Yoshida

摘要

We develope a pneumoconiosis classification model for chest X-ray images using Vision Transformer(ViT) and compare its accuracy with that of Convolutional Neural Networks (CNNs). Since there is variability in accuracy and differences in the accuracy across models, three ViT pretrained models are used for verification. Common CNN models, VGG16 and ResNet50, are used as comparison models. Then trials are performed for each training condition, and the average correct response rate is used as the evaluation index. The training results have shown that ViT is as accurate as or more accurate than CNN in all training conditions. This suggests that ViT is useful for pneumoconiosis classification. In terms of training time, ViT requires less time than CNN for a small number of epochs.