Enhancing Energy-Based Scatter Estimation Using Energy Spectra Modification in PET

Abstract

Energy-based scatter estimation methods have illustrated promising results in recent literature. Accurate estimation of energy probability density function of scattered photons (PDF-SC) is essential for precise scatter estimation and avoiding bias in reconstructed images. This article presents a novel method, referred to as energy spectra modification (ESM), to precisely estimate position-dependent local PDF-SC, which improves the accuracy of scatter estimation. ESM involves an iterative process to deblur local energy spectra, with the starting point constructed using an initial PDF-SC derived from global energy spectra. The scattered component of the deblurred energy spectrum is reblurred and normalized to estimate the local PDF-SC. We validated this approach through Monte Carlo simulations using a bladder phantom, an image quality phantom, and a cylindrical phantom. Comparative analyses were conducted against the traditional method employing global PDF-SC, a recent advancement, and the single scatter simulation method. The results demonstrated that our method effectively reduced activity bias of the global PDF-SC approach across various energy resolutions, windows, target size, and count levels. It achieved this with a comparable computational load and without hyperparameter modification.