Contrast Dependence of Spatial Resolution in Point Spread Function Reconstruction: A Simulation Study

抄録

This study simulated PSF reconstruction with edge artifacts mitigation while recovering spatial resolution under Poisson noise, and investigated the cause of contrast dependence. In 2D numerical phantoms consisted of disk 184 mm in diameter and different contrast of 7 disks 20 mm in diameter, we blurred the projection data using 1D Gaussian detector response with 5 mm FWHM. Projection matrix of parallel beam geometry considered only detector response and did not include other components so as to edge artifacts were not influenced. Reconstruction was performed using OSEM regularized with total variation (TV). To suppress noise and to perform stable deconvolution, blur functions with 9.94 mm FWHM given by the convolution of detector response 5 mm FWHM and 1D Gaussian filter with 8.59 mm FWHM was used in forward projection. FWHM of edge spread function was measured as a spatial resolution index. FWHM of PSF-TV reconstruction at projection data of 5 × 10⁷ counts decreased as 1.55, 1.29, 1.28, 1.17, 1.09, 1.11, 1.04 mm at contrast of 0.5, 1, 2, 3, 4, 5, 6, respectively. Contrast dependence of spatial resolution in the presence of noise is due to TV regularization, which acts to smooth objects with low-contrast more strongly than those with high-contrast.