Steerable Conditional Diffusion for Domain Adaptation in PET Image Reconstruction

2025-10-15

AI Summary

Diffusion models for PET image reconstruction suffer from domain shift—exhibiting hallucinatory artifacts when generalizing across anatomies, acquisition protocols, or pathological conditions. To address this, we propose a controllable conditional diffusion reconstruction framework that, for the first time, synergistically integrates steering conditional diffusion with low-rank adaptation (LoRA) for PET reconstruction. This enables online prior correction at each denoising step, dynamically aligning the model’s learned prior with the target-domain distribution. Built upon the PET-LiSch framework, our method is validated on synthetic 2D brain PET data, where it effectively suppresses out-of-distribution hallucinated structures. Quantitative and qualitative evaluations demonstrate superior reconstruction fidelity compared to OSEM and state-of-the-art diffusion-based methods, with significantly improved generalization and robustness under domain shifts.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Intelligent Robots: State Estimation

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Diffusion models have recently enabled state-of-the-art reconstruction of positron emission tomography (PET) images while requiring only image training data. However, domain shift remains a key concern for clinical adoption: priors trained on images from one anatomy, acquisition protocol or pathology may produce artefacts on out-of-distribution data. We propose integrating steerable conditional diffusion (SCD) with our previously-introduced likelihood-scheduled diffusion (PET-LiSch) framework to improve the alignment of the diffusion model's prior to the target subject. At reconstruction time, for each diffusion step, we use low-rank adaptation (LoRA) to align the diffusion model prior with the target domain on the fly. Experiments on realistic synthetic 2D brain phantoms demonstrate that our approach suppresses hallucinated artefacts under domain shift, i.e. when our diffusion model is trained on perturbed images and tested on normal anatomy, our approach suppresses the hallucinated structure, outperforming both OSEM and diffusion model baselines qualitatively and quantitatively. These results provide a proof-of-concept that steerable priors can mitigate domain shift in diffusion-based PET reconstruction and motivate future evaluation on real data.

Problem

• Addressing domain shift in PET image reconstruction
• Mitigating hallucinated artifacts from out-of-distribution data
• Aligning diffusion model priors with target subject anatomy

Innovation

• Steerable conditional diffusion aligns prior with target
• Low-rank adaptation dynamically adjusts model during reconstruction
• Likelihood-scheduled framework suppresses domain shift artefacts