Pixels simultaneous detection probabilities and spatial resolution determination of pixelized detectors by means of correlation measurements

TL;DR

This paper introduces a new correlation-based method to accurately estimate the detection probability and spatial resolution of pixelized detectors, validated through mammographic imaging and simulations, revealing complex PSF structures.

Contribution

The paper presents a novel correlation measurement approach for determining pixel detection probabilities and detailed PSF analysis, improving resolution estimation accuracy.

Findings

Pixel detection probabilities estimated within 0.001-0.003 accuracy.

Single Gaussian approximation insufficient for PSF; two-Gaussian model fits better.

Spatial resolution estimated at 54-58 micrometers for tested units.

Abstract

A novel method to estimate the pixels simultaneous detection probability and the spatial resolution of pixelized detectors is proposed, which is based on the determination of the statistical correlations between detector neighbor pixels. The correlations are determined by means of noise variance measurement for a isolated pixels and the difference between neighbor pixels. The method is validated using images from the two different GE Senographe 2000D mammographic units. The pixelized detector has been irradiated using x-rays along its entire surface. It is shown that the pixel simultaneous detection probabilities can be estimated within accuracy 0.001 - 0.003, where the systematic error is estimated to be smaller than 0.005. The presampled two-dimensional point-spread function (PSF0) is determined using a single Gaussian and a sum of two Gaussian approximations. The obtained results for the presampled PSF0 show that the single Gaussian approximation is not appropriate, and the sum of two Gaussian approximations providing the best fit predicts the existence of a large (~50%) narrow component. Another proof of this fact is the latest simulation study of columnar indirect digital detectors by A. Badano et al. The sampled two-dimensional PSF is determined using Monte Carlo simulation for the L-shape uniform distributed acceptance function for different values of fill factors. The detector spatial resolution is estimated using sampled PSF and has values 54 and 58 mkm for two different units. The calculation of the presampled modulation transfer function based on the PSF0 estimation shows that the existing data can only be reproduced using a single Gaussian approximation and the usage of the sum of two Gaussian show significantly larger values in the higher frequency region for both units.