AnF-DiffPET: Anatomy- and Frequency-Guided Diffusion for PET/CT Denoising

2026-07-01

AI Summary

This study addresses the poor quantitative reliability of low-dose PET imaging, which suffers from high noise levels, insufficient anatomical guidance, and unstable frequency-domain information recovery. To overcome these challenges, the authors propose a diffusion model that integrates CT-derived anatomical priors with frequency-domain guidance through an anatomy–frequency collaborative mechanism for high-quality denoising. Key innovations include an Anatomy–Frequency Guidance (AFG) strategy, a Multi-Scale Cross-Transformer Reconstruction (MSCTR) module, and Frequency-domain Contrastive Hard Example Mining (FCHM), collectively enhancing both anatomical consistency and spectral fidelity. Extensive experiments on four PET/CT datasets demonstrate that the proposed method consistently outperforms state-of-the-art approaches—including CNNs, GANs, Transformers, and existing diffusion models—in terms of image fidelity, anatomical alignment, and quantitative accuracy.

Abstract

Positron emission tomography (PET) provides essential functional information for disease assessment, however reducing injected activity or acquisition time produces low-dose (LD) PET with stronger count dependent noise and less reliable uptake quantification. Diffusion models offer a promising solution for PET denoising by progressively recovering high-dose (HD) PET images from LD inputs. However, LD-to-HD PET denoising is still challenging due to insufficient anatomical guidance, unstable multi-scale feature propagation, and uncertain frequency domain uptake recovery. We propose AnF-DiffPET, an anatomy- and frequency-guided diffusion framework for computed tomography (CT) conditioned LD PET denoising. The framework integrates Anatomical-Frequency Guidance (AFG), Multi-Scale Cross-Transformer Reconstruction (MSCTR), and Frequency-Contrastive Hard Mining (FCHM) to enhance anatomy aware feature modulation and frequency domain consistency during denoising. Experimental results across four PET/CT datasets show that the proposed method improves image fidelity, anatomical consistency, and quantitative fidelity over representative CNN-based, GAN-based, transformer-based, and diffusion-based methods. The code and trained models will be publicly released upon acceptance.

Problem

• PET denoising
• low-dose PET
• anatomical guidance
• frequency domain recovery
• image fidelity

Innovation

• Anatomy-guided diffusion
• Frequency-domain denoising
• Multi-scale Cross-Transformer
• PET/CT image reconstruction
• Low-dose PET