Affiliations: Royal Society Wolfson Bioinformatics Research
Laboratory, School of Computing Sciences, University of East Anglia, Norwich,
UK | School of Biological Sciences, and School of Chemical
Sciences and Pharmacy, University of East Anglia, Norwich, UK | School of Computing Sciences, University of East
Anglia, Norwich, UK
Note: [] Corresponding author: Dr. M. Razaz, Royal Society Wolfson
Bioinformatics Research Lab, School of Computing Sciences, University of East
Anglia, Norwich NR4 7TJ, UK. E-mail: m.razaz@uea.ac.uk
Abstract: X-ray diffraction imaging is an important analytical tool in a
variety of biological and medical imaging applications. However, the images
obtained are often degraded by the point spread function (PSF) of the imaging
system. An accurate knowledge of the system PSF is essential for a successful
deconvolution of the degraded images. In this paper, we present a novel
modeling procedure to determine the PSF of an X-ray diffraction (XRD) system
based on experimental data. Different regions of an XRD pattern have different
orientations due to the diffuse light distortion (DLD). A new multiple PSF
model is introduced and used to restore XRD data. Raw PSFs are collected using
isolated spots from XRD data in high resolution regions. An adaptive ridge
regression (ARR) technique is first used to remove noise and baseline from the
raw PSF data. A target Gaussian function is used to model the raw PSFs. A
gradient descent algorithm is used to find optimum parameters of the Gaussian
function. A set of XRD data are restored by a nonlinear iterative deconvolution
algorithm using the modeled PSFs. Experimental results using a single and
multiple PSFs are presented and discussed. We show that by using a multiple PSF
model in the deconvolution algorithm, improved restored XRD data are obtained
and as a result the symmetry estimator and integration error are enhanced.
