AnF-DiffPET: anatomy- and frequency-guided diffusion for simulated low-dose PET/CT denoising.
Authors
Affiliations (5)
Affiliations (5)
- Northeastern University, Northeastern University, Shenyang, China, Shenyang, 110819, China.
- Tohoku University School of Medicine, Tohoku University School of Medicine, Sendai, 980-8575, Japan.
- Northeastern University, Northeastern University, Shenyang, China, Shenyang, Liaoning, 110819, China.
- University of Dundee, DD1 4HN, Dundee, Scotland, DD1 4HN, United Kingdom of Great Britain and Northern Ireland.
- College of Medicine and Biological Information Engineering, Northeastern University, Northeastern University, Shenyang, China Shenyang, CN 110819, Shenyang, 110819, China.
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 remains challenging because of limited anatomical guidance, unstable multi-scale feature propagation, and inaccurate recovery of uptake patterns in the frequency domain.

Approach: 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.

Main results: Experimental results across four simulated low-dose PET/CT datasets show that the proposed method improves image fidelity, anatomical consistency, and quantitative fidelity over representative CNN-based, GAN-based, Transformer-based, Mamba-based, and diffusion-based methods.
Significance: AnF-DiffPET provides a PET/CT-specific diffusion restoration framework that combines anatomical guidance, multi-scale feature reconstruction, and frequency-domain regularization for LD PET denoising. The code and trained models will be publicly released upon acceptance.