Less Is More: Training-Free Acceleration Framework of 3D Diffusion Models for Low-Count PET Denoising via Global-Local Trajectory Reduction

2026-06-07

AI Summary

This work addresses the high inference latency of diffusion model–based methods in low-count PET image reconstruction, which stems from iterative 3D volumetric sampling and hinders clinical deployment. The authors propose a training-free global-local skip-step strategy that accelerates the reverse diffusion process without modifying or retraining the pretrained diffusion model. By initializing intermediate denoising states and reusing high-level U-Net features from neighboring timesteps, the method achieves substantial speedup while preserving image fidelity. This approach introduces the first training-free trajectory compression mechanism with plug-and-play compatibility, yielding over 10× acceleration across multiple PET tracers. Reconstructed images match or exceed the quality of full-step baselines, and blinded reader studies confirm improved diagnostic confidence and perceptual image quality.

Abstract

Accurate quantification and uptake measurement in PET are critical for assessing disease progression and supporting clinical decision-making. While high-count PET provides reliable image quality, the associated radiation dose and prolonged acquisition remain significant clinical concerns, motivating the adoption of low-count protocols. Diffusion-model-based methods have demonstrated strong potential for restoring low-count PET to near high-count quality, but their iterative sampling procedure becomes prohibitively expensive when applied to high-resolution 3D PET volumes, introducing substantial inference latency that limits practical clinical deployment. To address these challenges, we propose a training-free Global-Local Skipping Strategy that accelerates diffusion model-based 3D PET denoising while simultaneously improving reconstruction quality. The proposed method is plug-and-play and directly applicable to pre-trained diffusion models without retraining or architectural modification. Specifically, we introduce: (i) a global denoising step skipping strategy that initializes the reverse diffusion process from an intermediate denoising step using a noise-consistent transformation of the low-count input, substantially reducing the number of required denoising steps; and (ii) a local feature reuse shortcut that reuses slowly-varying high-level U-Net features across neighboring denoising steps, further reducing per-step computation while preserving image fidelity. We evaluate the proposed approach on multiple PET tracers from in-house and public datasets, including 18F-FDG PET, 68Ga-DOTATATE PET, and 18F-PSMA PET, demonstrating consistent acceleration of over an order of magnitude alongside improved or comparable reconstruction performance relative to the full-step baseline. Blinded reader studies further confirm enhanced clinical confidence and perceived diagnostic quality.

Problem

• PET denoising
• 3D diffusion models
• low-count PET
• inference latency
• image reconstruction

Innovation

• training-free acceleration
• 3D diffusion models
• PET denoising
• global-local skipping
• feature reuse