Scholarlink

This research aims to denoise low-dose computed tomography images using a U-Net rather than the conventional ways of denoising images including spatial filtering, wavelet thresholding, and transform domain filtering. An effi- cient denoising model should reduce noise while keeping edges intact. CNNs produce superior outcomes as compared to older approaches while also being far more efficient in their use of computing resources. This research used a method of stacking a multilayer convolution layer on top of a batch normalization (BN) layer and an activation function layer. This improved learning module maintains the image's finer details and successfully removes noise from CT images while maintaining image quality by using peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) as evaluation metrics. The Mayo Clinic LDCT Grand Challenge dataset obtained a PSNR of 43.106 and SSIM of 0.972.

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

U-Net Architecture for Denoising Low Dose CT Scanned Images

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Swati Chauhan,
  • Rekha Vig,
  • Nidhi Malik

摘要

This research aims to denoise low-dose computed tomography images using a U-Net rather than the conventional ways of denoising images including spatial filtering, wavelet thresholding, and transform domain filtering. An effi- cient denoising model should reduce noise while keeping edges intact. CNNs produce superior outcomes as compared to older approaches while also being far more efficient in their use of computing resources. This research used a method of stacking a multilayer convolution layer on top of a batch normalization (BN) layer and an activation function layer. This improved learning module maintains the image's finer details and successfully removes noise from CT images while maintaining image quality by using peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) as evaluation metrics. The Mayo Clinic LDCT Grand Challenge dataset obtained a PSNR of 43.106 and SSIM of 0.972.