Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PatchDenoiser: Parameter-efficient multi-scale patch learning and fusion denoiser for medical images

Jitindra Fartiyal, Pedro Freire, Sergei K. Turitsyn, Sergei G. Solovski

TL;DR

PatchDenoiser is a lightweight, energy-efficient multi-scale patch-based denoising framework that decomposes denoising into local texture extraction and global context aggregation, fused via a spatially aware patch fusion strategy that enables effective noise suppression while preserving fine structural and anatomical details.

Abstract

Medical images are essential for diagnosis, treatment planning, and research, but their quality is often degraded by noise from low-dose acquisition, patient motion, or scanner limitations, affecting both clinical interpretation and downstream analysis. Traditional filtering approaches often over-smooth and lose fine anatomical details, while deep learning methods, including CNNs, GANs, and transformers, may struggle to preserve such details or require large, computationally expensive models, limiting clinical practicality. We propose PatchDenoiser, a lightweight, energy-efficient multi-scale patch-based denoising framework. It decomposes denoising into local texture extraction and global context aggregation, fused via a spatially aware patch fusion strategy. This design enables effective noise suppression while preserving fine structural and anatomical details. PatchDenoiser is ultra-lightweight, with far fewer parameters and lower computational complexity than CNN-, GAN-, and transformer-based denoisers. On the 2016 Mayo Low-Dose CT dataset, PatchDenoiser consistently outperforms state-of-the-art CNN- and GAN-based methods in PSNR and SSIM. It is robust to variations in slice thickness, reconstruction kernels, and HU windows, generalizes across scanners without fine-tuning, and reduces parameters by ~9x and energy consumption per inference by ~27x compared with conventional CNN denoisers. PatchDenoiser thus provides a practical, scalable, and computationally efficient solution for medical image denoising, balancing performance, robustness, and clinical deployability.

PatchDenoiser: Parameter-efficient multi-scale patch learning and fusion denoiser for medical images

TL;DR

PatchDenoiser is a lightweight, energy-efficient multi-scale patch-based denoising framework that decomposes denoising into local texture extraction and global context aggregation, fused via a spatially aware patch fusion strategy that enables effective noise suppression while preserving fine structural and anatomical details.

Abstract

Medical images are essential for diagnosis, treatment planning, and research, but their quality is often degraded by noise from low-dose acquisition, patient motion, or scanner limitations, affecting both clinical interpretation and downstream analysis. Traditional filtering approaches often over-smooth and lose fine anatomical details, while deep learning methods, including CNNs, GANs, and transformers, may struggle to preserve such details or require large, computationally expensive models, limiting clinical practicality. We propose PatchDenoiser, a lightweight, energy-efficient multi-scale patch-based denoising framework. It decomposes denoising into local texture extraction and global context aggregation, fused via a spatially aware patch fusion strategy. This design enables effective noise suppression while preserving fine structural and anatomical details. PatchDenoiser is ultra-lightweight, with far fewer parameters and lower computational complexity than CNN-, GAN-, and transformer-based denoisers. On the 2016 Mayo Low-Dose CT dataset, PatchDenoiser consistently outperforms state-of-the-art CNN- and GAN-based methods in PSNR and SSIM. It is robust to variations in slice thickness, reconstruction kernels, and HU windows, generalizes across scanners without fine-tuning, and reduces parameters by ~9x and energy consumption per inference by ~27x compared with conventional CNN denoisers. PatchDenoiser thus provides a practical, scalable, and computationally efficient solution for medical image denoising, balancing performance, robustness, and clinical deployability.
Paper Structure (17 sections, 2 equations, 4 figures, 6 tables)

This paper contains 17 sections, 2 equations, 4 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Method
  4. Patch Feature Extractor (PFE)
  5. Patch Fusion Module (PFM)
  6. Patch Consolidator Module (PCM)
  7. Experimental Setup
  8. Datasets
  9. Preprocessing
  10. Training
  11. Evaluation Metrics
  12. Results
  13. Evaluation on Variations in Hounsfield Unit (HU) Window, Slice Thickness, and Reconstruction Kernel
  14. Cross-Dataset Generalization
  15. Ablation Studies
  16. ...and 2 more sections

Figures (4)

  • Figure 1: Overview of the PatchDenoiser framework showing the three main modules: Patch Feature Extractor (PFE), Patch Fusion Module (PFM), and Patch Consolidator Module (PCM).
  • Figure 2: Illustration of the Patch Fusion Module (PFM), where multi-scale feature maps are combined using a gated fusion mechanism.
  • Figure 3: Qualitative comparison of denoised abdomen LDCT images on the 2016 Mayo dataset. Zoomed-in regions of interest highlight differences in noise suppression and structural detail preservation among the evaluated methods.
  • Figure 4: Qualitative denoising results under different slice thicknesses and reconstruction kernels on the 2016 Mayo LDCT dataset.