Fast and accurate generation method of PSF-based system matrix for PET reconstruction

TL;DR

This paper introduces a fast, accurate method based on P-SPIR theory for generating PSF-contained system matrices in PET reconstruction, improving resolution and reducing computational costs with GPU acceleration.

Contribution

The paper presents a novel P-SPIR-based approach that considers both incidence angle and position, enabling rapid and precise system matrix generation for PET imaging.

Findings

Enhanced spatial resolution in PET images.

Rapid system matrix generation within less than 1 second.

Effective suppression of depth of interaction effects.

Abstract

Positional single photon incidence response (P-SPIR) theory is researched in this paper to generate more accurate PSF-contained system matrix simply and quickly. The method has been proved highly effective to improve the spatial resolution by applying to the Eplus-260 primate PET designed by the Institute of High Energy Physics of the Chinese Academy of Sciences(IHEP). Simultaneously, to meet the clinical needs, GPU acceleration is put to use. Basically, P-SPIR theory takes both incidence angle and incidence position by crystal subdivision instead of only incidence angle into consideration based on Geant4 Application for Emission Tomography (GATE). The simulation conforms to the actual response distribution and can be completed rapidly within less than 1s. Furthermore,two-block penetration and normalization of the response probability are raised to fit the reality. With PSF obtained, the homogenization model is analyzed to calculate the spread distribution of bins within a few minutes for system matrix generation. As a reult, The images reconstructed indicate that the P-SPIR method can effectively inhibit the depth of interaction (DOI) effect especially in the field close to the edge of the field of view (FOV). What is more, the method can be promoted to any other PET and the list-mode organization structure high-speedily and efficiently, which substantially reduces the computing cost and ensures the accuracy of system matrix for PET reconstruction.