Likelihood-Scheduled Score-Based Generative Modeling for Fully 3D PET Image Reconstruction

TL;DR

This paper introduces a likelihood-scheduled score-based generative modeling approach for fully 3D PET image reconstruction, significantly improving speed, consistency, and hyperparameter tuning over existing methods, and demonstrating effectiveness on real 3D PET data.

Contribution

It proposes a novel likelihood-matching methodology for 3D PET reconstruction that accelerates process and reduces hyperparameter tuning, with successful application to real 3D PET data.

Findings

Matches or improves NRMSE and SSIM compared to state-of-the-art methods

Reduces reconstruction time and hyperparameter tuning requirements

First implementation of SGM-based reconstruction for real 3D PET data

Abstract

Medical image reconstruction with pre-trained score-based generative models (SGMs) has advantages over other existing state-of-the-art deep-learned reconstruction methods, including improved resilience to different scanner setups and advanced image distribution modeling. SGM-based reconstruction has recently been applied to simulated positron emission tomography (PET) datasets, showing improved contrast recovery for out-of-distribution lesions relative to the state-of-the-art. However, existing methods for SGM-based reconstruction from PET data suffer from slow reconstruction, burdensome hyperparameter tuning and slice inconsistency effects (in 3D). In this work, we propose a practical methodology for fully 3D reconstruction that accelerates reconstruction and reduces the number of critical hyperparameters by matching the likelihood of an SGM's reverse diffusion process to a current iterate of the maximum-likelihood expectation maximization algorithm. Using the example of low-count reconstruction from simulated [$^{18}$F]DPA-714 datasets, we show our methodology can match or improve on the NRMSE and SSIM of existing state-of-the-art SGM-based PET reconstruction while reducing reconstruction time and the need for hyperparameter tuning. We evaluate our methodology against state-of-the-art supervised and conventional reconstruction algorithms. Finally, we demonstrate a first-ever implementation of SGM-based reconstruction for real 3D PET data, specifically [$^{18}$F]DPA-714 data, where we integrate perpendicular pre-trained SGMs to eliminate slice inconsistency issues.