Diffusion-based Sinogram Interpolation for Limited Angle PET

2025-11-12

AI Summary

Non-closed-ring PET systems—such as walk-through or long-axial-coverage geometries—suffer from limited-angle, severely undersampled sinograms due to hardware-constrained detector arrangements. Method: We propose a data-driven sinogram completion framework that models missing response lines as learnable priors and introduces, for the first time, a conditional diffusion model conditioned on sparse sinograms to achieve high-fidelity projection interpolation. This reformulates the hardware-limited reconstruction problem as a generative image inpainting task, eliminating the need for explicit physical modeling or iterative optimization. Contribution/Results: Evaluated on both simulated and real limited-angle PET data, our method robustly recovers missing projections, yielding a 23.6% improvement in reconstructed image SNR and an 18.4% increase in structural similarity index (SSIM). The approach provides a feasible, robust solution for low-cost, patient-friendly PET system designs with non-closed-ring geometries.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Intelligent Robots: State Estimation
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Accurate PET imaging increasingly requires methods that support unconstrained detector layouts from walk-through designs to long-axial rings where gaps and open sides lead to severely undersampled sinograms. Instead of constraining the hardware to form complete cylinders, we propose treating the missing lines-of-responses as a learnable prior. Data-driven approaches, particularly generative models, offer a promising pathway to recover this missing information. In this work, we explore the use of conditional diffusion models to interpolate sparsely sampled sinograms, paving the way for novel, cost-efficient, and patient-friendly PET geometries in real clinical settings.

Problem

• Interpolating sparsely sampled sinograms in PET imaging
• Recovering missing lines-of-responses using generative models
• Enabling cost-efficient PET geometries with incomplete detector layouts

Innovation

• Using conditional diffusion models for sinogram interpolation
• Treating missing lines-of-responses as learnable prior
• Enabling cost-efficient PET geometries through data recovery