Evaluating the Posterior Sampling Ability of Plug&Play Diffusion Methods in Sparse-View CT

2024-10-21

IEEE International Conference on Acoustics, Speech, and Signal Processing

AI Summary

This work addresses the posterior sampling reliability of Plug-and-Play (PnP) diffusion models in sparse-view CT reconstruction, systematically evaluating their approximation fidelity to the true posterior distribution under extremely low projection counts. We introduce two novel quantitative metrics for posterior fidelity assessment and conduct a comparative evaluation across three CT datasets among PnP-DDRM, PnP-DPS, and PnP-CDM. Results show that posterior approximation quality degrades sharply as the number of projections decreases; high PSNR/SSIM reconstruction scores do not guarantee posterior consistency. All existing PnP diffusion frameworks exhibit significant deviation from the true posterior—particularly failing in multimodal or flat posterior regimes. This study exposes a fundamental limitation of PnP diffusion models in uncertainty quantification, highlighting critical challenges for trustworthy Bayesian reconstruction in medical imaging and establishing essential evaluation benchmarks for future methodological development.

Technology Category

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

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Web Mining and Content Analysis: Web data provenance, reliability, and authenticity
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Plug&Play (PnP) diffusion models are state-of-the-art methods in computed tomography (CT) reconstruction. Such methods usually consider applications where the sinogram contains a sufficient amount of information for the posterior distribution to be peaked, and consequently are evaluated using image-to-image metrics such as PSNR/SSIM. Instead, we are interested in reconstructing compressible flow images from sinograms having a small number of projections, which results in a posterior distribution no longer peaked or even multimodal. Thus, in this paper, we aim at evaluating the approximate posterior of PnP diffusion models and introduce two posterior evaluation criteria. We quantitatively evaluate three PnP diffusion methods on three different datasets for several numbers of projections. We surprisingly find that, for each method, the approximate posterior deviates from the true posterior when the number of projections decreases.

Problem

• Evaluates Plug&Play diffusion models in sparse-view CT reconstruction.
• Focuses on reconstructing compressible flow images from limited projections.
• Assesses deviation of approximate posterior from true posterior.

Innovation

• Evaluates Plug&Play diffusion models in sparse-view CT
• Introduces two posterior evaluation properties
• Quantitatively assesses three PnP methods across datasets