Abstract:Low-dose computed tomography (LDCT) has emerged as the preferred technology for diagnostic medical imaging due to the potential health risks associated with X-ray radiation and conventional computed tomography (CT) techniques. While LDCT utilizes a lower radiation dose compared to standard CT, it results in increased image noise, which can impair the accuracy of diagnoses. To mitigate this issue, advanced deep learning-based LDCT denoising algorithms have been developed. These primarily utilize Convolutional Neural Networks (CNNs) or Transformer Networks and often employ the Unet architecture, which enhances image detail by integrating feature maps from the encoder and decoder via skip connections. However, existing methods focus excessively on the optimization of the encoder and decoder structures while overlooking potential enhancements to the Unet architecture itself. This oversight can be problematic due to significant differences in feature map characteristics between the encoder and decoder, where simple fusion strategies may hinder effective image reconstruction. In this paper, we introduce WiTUnet, a novel LDCT image denoising method that utilizes nested, dense skip pathway in place of traditional skip connections to improve feature integration. Additionally, to address the high computational demands of conventional Transformers on large images, WiTUnet incorporates a windowed Transformer structure that processes images in smaller, non-overlapping segments, significantly reducing computational load. Moreover, our approach includes a Local Image Perception Enhancement (LiPe) module within both the encoder and decoder to replace the standard multi-layer perceptron (MLP) in Transformers, thereby improving the capture and representation of local image features. Through extensive experimental comparisons, WiTUnet has demonstrated superior performance over existing methods in critical metrics such as Peak Signal-to-Noise Ratio (PSNR), Structural Similarity (SSIM), and Root Mean Square Error (RMSE), significantly enhancing noise removal and image quality. The code is available on github https://github.com/woldier/WiTUNet.

Spectrally Consistent UNet for High Fidelity Image Transformations

U-Net-Based Multispectral Image Generation from an RGB Image

TransUNetCD: A Hybrid Transformer Network for Change Detection in Optical Remote-Sensing Images

WiTUnet: A U-shaped architecture integrating CNN and Transformer for improved feature alignment and local information fusion

Fast and Accurate Single Image Super- Resolution Via Enhanced U-Net

Scale-Equivariant UNet for Histopathology Image Segmentation

FAN-Unet: Enhancing Unet with vision Fourier Analysis Block for Biomedical Image Segmentation

Enhanced Deep Unrolling Networks for Snapshot Compressive Hyperspectral Imaging

TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation

Hyperspectral Image Denoising via Spatial–Spectral Recurrent Transformer

Hyperspectral Image Denoising via Spatial-Spectral Recurrent Transformer

ACC-UNet: A Completely Convolutional UNet model for the 2020s

Focal-UNet: UNet-like Focal Modulation for Medical Image Segmentation

Index Network

Multiscale image denoising algorithm based on UNet3+

Optimization-inspired Cumulative Transmission Network for image compressive sensing

ColorUNet: A convolutional classification approach to colorization

Convolutional Neural Networks on non-uniform geometrical signals using Euclidean spectral transformation

GuidedNet: A General CNN Fusion Framework via High-Resolution Guidance for Hyperspectral Image Super-Resolution

Unmixing Guided Unsupervised Network for RGB Spectral Super-Resolution