Evaluation of the point spread function for restoration of autoradiographic images

This thesis is concerned with the resolution enhancement of images generated by autoradiography. The approach taken is to accurately describe the imaging process of autoradiography using a mathematical model and to use that model to experimentally measure the point spread function (PSF) and employ image restoration algorithms in order to improve the resolution. A complex nonlinear model, which describes the imaging process, is developed, analyzed and experimentally validated. Several simplifications of this model are considered and two smaller models are investigated: a linear shift invariant model with independent additive Gaussian noise and a signal-dependent film grain noise model implemented in the optical density domain. The point spread function of autoradiography is experimentally measured using {dollar}sp{lcub}51{rcub}{dollar}Cr microspheres with an average diameter of 10 {dollar}mu{dollar}m. The observed response is fitted to a theoretical expression for the point spread function of autoradiography using a nonlinear least squares regression algorithm based on Marquardt's method. Wiener filters and adaptive iterative regularized image restoration algorithms are applied to the autoradiographs generated by the microspheres. Resolution measurements are quantified in terms of the Full-Width at Half-Maximum (FWHM) criterion.; The experimental results demonstrate that the iterative algorithms developed for the signal-independent model increase the FWHM resolution from 184 {dollar}mu{dollar}m to approximately 134 {dollar}mu{dollar}m, which represents a relative improvement of about 27%. For the signal-dependent model, a weighted iterative adaptive regularization algorithm that incorporates the signal-dependent nature of the film grain noise provides a 43% improvement in the FWHM resolution to nearly 105 {dollar}mu{dollar}m. Application of these image restoration algorithms is extended to autoradiographic experiments of kidneys using {dollar}sp{lcub}51{rcub}{dollar}Cr labeled EDTA and {dollar}sp{lcub}51{rcub}{dollar}Cr labeled red blood cells. The PSF from the nonlinear least squares regression is adjusted for tissue thickness and the results indicate an effective increase in resolution and a suppression of background film grain noise. The use of these image restoration techniques with other isotopes and other applications of autoradiography is discussed.