Supervised Diffusion-Model-Based PET Image Reconstruction

2025-06-30

AI Summary

Existing unsupervised diffusion models (DMs) for PET reconstruction fail to explicitly model the interaction between learned priors and Poisson-distributed measurement data, limiting reconstruction accuracy and uncertainty quantification—particularly under low-dose conditions. Method: We propose the first supervised diffusion prior framework: a DM is integrated as a learnable regularizer jointly optimized with the Poisson likelihood, thereby explicitly coupling the prior with the physical measurement model; we further introduce non-negativity constraints and intensity-adaptive output mechanisms to enable robust 3D full-volume reconstruction. Results: Evaluated on real FDG brain PET data across multiple dose levels, our method achieves superior or competitive reconstruction fidelity and uncertainty calibration compared to state-of-the-art methods. It significantly improves posterior sampling stability and enhances clinical applicability, demonstrating both quantitative accuracy and reliable uncertainty estimation in low-count regimes.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Diffusion Models for Vision
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Diffusion models (DMs) have recently been introduced as a regularizing prior for PET image reconstruction, integrating DMs trained on high-quality PET images with unsupervised schemes that condition on measured data. While these approaches have potential generalization advantages due to their independence from the scanner geometry and the injected activity level, they forgo the opportunity to explicitly model the interaction between the DM prior and noisy measurement data, potentially limiting reconstruction accuracy. To address this, we propose a supervised DM-based algorithm for PET reconstruction. Our method enforces the non-negativity of PET's Poisson likelihood model and accommodates the wide intensity range of PET images. Through experiments on realistic brain PET phantoms, we demonstrate that our approach outperforms or matches state-of-the-art deep learning-based methods quantitatively across a range of dose levels. We further conduct ablation studies to demonstrate the benefits of the proposed components in our model, as well as its dependence on training data, parameter count, and number of diffusion steps. Additionally, we show that our approach enables more accurate posterior sampling than unsupervised DM-based methods, suggesting improved uncertainty estimation. Finally, we extend our methodology to a practical approach for fully 3D PET and present example results from real [$^{18}$F]FDG brain PET data.

Problem

• Improves PET image reconstruction accuracy using supervised diffusion models
• Addresses limitations of unsupervised schemes in modeling noise-data interaction
• Enhances uncertainty estimation and handles wide PET intensity ranges

Innovation

• Supervised diffusion-model-based PET reconstruction
• Enforces non-negativity and wide intensity range
• Outperforms deep learning methods across doses