Scholarlink

Data augmentation enhances medical imaging tasks but faces domain gaps and fragmented studies. We propose a unified framework applying six mix-based methods on brain tumor MRI and eye disease datasets with convolutional and transformer backbones. Our contributions are threefold. (1) We present MediAug, a benchmark for advanced data augmentation in medical imaging. (2) Six methods (MixUp, YOCO, CropMix, CutMix, AugMix, SnapMix) are evaluated with ResNet-50 and ViT-B backbones. (3) Experiments show MixUp achieves 79.19% accuracy on brain tumor classification with ResNet-50, SnapMix 99.44% with ViT-B, YOCO 91.60% on eye disease classification with ResNet-50, and CutMix 97.94% with ViT-B. Code will be available at https://github.com/AIGeeksGroup/MediAug .

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

MediAug: Exploring Visual Augmentation in Medical Imaging

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Xuyin Qi,
  • Zeyu Zhang,
  • Canxuan Gang,
  • Hao Zhang,
  • Lei Zhang,
  • Zhiwei Zhang,
  • Yang Zhao

摘要

Data augmentation enhances medical imaging tasks but faces domain gaps and fragmented studies. We propose a unified framework applying six mix-based methods on brain tumor MRI and eye disease datasets with convolutional and transformer backbones. Our contributions are threefold. (1) We present MediAug, a benchmark for advanced data augmentation in medical imaging. (2) Six methods (MixUp, YOCO, CropMix, CutMix, AugMix, SnapMix) are evaluated with ResNet-50 and ViT-B backbones. (3) Experiments show MixUp achieves 79.19% accuracy on brain tumor classification with ResNet-50, SnapMix 99.44% with ViT-B, YOCO 91.60% on eye disease classification with ResNet-50, and CutMix 97.94% with ViT-B. Code will be available at https://github.com/AIGeeksGroup/MediAug .