2024 年 42 巻 4 号 p. 117-129
This study aimed to elucidate the cause of object-dependent spatial resolution of ordered subset expectation maximization (OSEM) regularized with edge-preserving potential prior based on the feature values such as Peak, Area, and Peak/Area of line spread function (LSF) using computer simulation. The spatial resolution was assessed using numerical phantoms of point source and disk source with contrast (signal/background−1) varying 0.01−9 with 11 steps. The detector response was assumed to be a one-dimensional (1D) Gaussian function with a shift-in variant 3 mm full width at half maximum (FWHM), and ideal 2D parallel beam data were used with no degrada tion factors other than blurring. Reconstruction was performed using OSEM-RD (OSEM regularized by the rela tive difference prior (RD)) with 64 subsets and 20 iterations. Point source phantom used a 1D Gaussian function to approximate the LSF, whereas disk source phantom used the cumulative distribution function of a Gaussian function to approximate edge spread function (ESF). While the evaluation of spatial resolution differs, the corre spondence between the LSF and ESF formulas allows us to discuss spatial resolution using unified feature values between them. The FWHM of OSEM-RD for a disk source was smaller than that for a point source as expected by feature values.
