Scholarlink

<p>Computed Tomography (CT) involves harmful radiation, while Low-Dose CT (LDCT) reduces exposure but adds noise that can hinder diagnosis. To improve LDCT image quality, we fine-tuned four segmentation models with six pretrained encoders and evaluated them using PSNR and SSIM. Experiments on the Mayo Clinic dataset showed that U-Net with Inceptionv4 achieved the best performance (PSNR 48.797, SSIM 0.977), effectively reducing noise and preserving structural details.</p>

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

Evaluating multiple combinations of models and encoders to enhance LDCT images

  • 期刊论文

  • 发表日期：

    2025年7月23日

查看全文
  • Natalia Camillo do Carmo,
  • Larissa Ferreira Rodrigues Moreira,
  • André Ricardo Backes

摘要

Computed Tomography (CT) involves harmful radiation, while Low-Dose CT (LDCT) reduces exposure but adds noise that can hinder diagnosis. To improve LDCT image quality, we fine-tuned four segmentation models with six pretrained encoders and evaluated them using PSNR and SSIM. Experiments on the Mayo Clinic dataset showed that U-Net with Inceptionv4 achieved the best performance (PSNR 48.797, SSIM 0.977), effectively reducing noise and preserving structural details.