Pseudo-MRI-Guided PET Image Reconstruction Method Based on a Diffusion Probabilistic Model

TL;DR

A diffusion probabilistic model is employed to infer T1-weighted-MRI (deep-MRI) images from FDG-PET brain images from FDG-PET brain images, enabling the possibility of changing reconstruction parameters, such as the strength of the prior on anatomically guided PET reconstruction in the absence of MRI.

Abstract

Anatomically guided PET reconstruction using MRI information has been shown to have the potential to improve PET image quality. However, these improvements are limited to PET scans with paired MRI information. In this work we employed a diffusion probabilistic model (DPM) to infer T1-weighted-MRI (deep-MRI) images from FDG-PET brain images. We then use the DPM-generated T1w-MRI to guide the PET reconstruction. The model was trained with brain FDG scans, and tested in datasets containing multiple levels of counts. Deep-MRI images appeared somewhat degraded than the acquired MRI images. Regarding PET image quality, volume of interest analysis in different brain regions showed that both PET reconstructed images using the acquired and the deep-MRI images improved image quality compared to OSEM. Same conclusions were found analysing the decimated datasets. A subjective evaluation performed by two physicians confirmed that OSEM scored consistently worse than the MRI-guided PET images and no significant differences were observed between the MRI-guided PET images. This proof of concept shows that it is possible to infer DPM-based MRI imagery to guide the PET reconstruction, enabling the possibility of changing reconstruction parameters such as the strength of the prior on anatomically guided PET reconstruction in the absence of MRI.

Figures (8)

• Figure 1: An illustration of the proposed deep-MRI-guided PET image reconstruction. Firstly, we generate a deep-MRI image from the OSEM PET image by using a pre-trained diffusion model (see its training process in Section \ref{['sec:DPM']} and Section \ref{['sec:cDPM']}). Then, the deep-MRI-guided PET image is obtained by solving an optimization of the anatomically-guided reconstruction given the generated deep-MRI images and the raw PET data.
• Figure 2: MRI acquired from scanner (top left), deep-MRI (top center), relative difference (top right), reconstructed PET image using OSEM (bottom left), gray and white matter masks (bottom center) and histogram of gray and white matter from acquired and deep-MRI images (bottom right) of an exemplar subject.
• Figure 3: Axial slices at different levels of reconstructed PET image using OSEM (first column), MRIq-PET image obtained using the acquired MRI (second column), MRIg-PET image obtained using the deep-MRI (third column), and relative difference between the MRIg-PET images obtained with the acquired and deep-MRI images.
• Figure 4: Mean relative difference between MRIg-PET reconstruction using the acquired and deep-MRI images and between MRIg-PET reconstruction using the the acquired MRI and OSEM for each ROI, averaged between eight subjects (top left). CV measured in each ROI obtained with MRIg-PET reconstruction using the acquired and deep-MRI images, and OSEM, averaged between eight subjects (bottom left).
• Figure 5: Axial slices of PET reconstructed images from the full dataset (first row), 25% of the counts (second row), 5% of the counts (third row), reconstructed with OSEM (left column), and with MRIg-PET using the acquired MRI (second column) and the deep-MRI (third column). The relative difference between the PET reconstructed images using MRIg-PET with the measured and deep-MRI are shown in the fourth column.