PubMed Health⌕ Search

Biomedical subjects

R G Paxman

Publications and source records attributed to R G Paxman.

3 recordsLinked to original sources

Image reconstruction from coded data: I. Reconstruction algorithms and experimental results.

Two algorithms have been developed for reconstructing objects from their coded images and a priori knowledge of the object class. Reconstructions from both algorithms are presented, but the results appear to be largely independent of the algorithm used. One of the algorithms, a Monte Carlo approach, is used to investigate the quality of the reconstruction of two- and three-dimensional objects from simulated coded-image data with respect to viewing geometry and multiplexing (mixing) of the data. The cases examined include reconstructions from data with and without signal-dependent photon noise. It is found that reconstructing from multiplexed data is not so serious a problem as reconstructing from data obtained with a limited viewing angle. Also, when photon noise is included in the data, reconstructions obtained from multiplexed data are better than those obtained from unmultiplexed data because of the higher photon count made available by multiplexing. It appears that the fidelity of a reconstruction depends much more strongly on the design of the data-taking system (the coded apertures) than on the reconstruction algorithm.

Models, Structural↗

Image reconstruction from coded data: II. Code design.

A strategy is given for the design of coded apertures with respect to a given class of objects that are to be imaged. Previous knowledge of the first- and second-order statistics for the object class is assumed. The object class is characterized by its Karhunen-Loève eigenvectors and eigenvalues, whereas the imaging system is characterized by its singular-value decomposition. We introduce the concept of alignment in which the aperture parameters are adjusted until the system is tuned to measure the given object class well. A mean-square-error figure of merit that indicates degree of alignment is given, and alignment is performed by standard optimization techniques. We illustrate this technique with a simple proof-of-principle experiment. These concepts are general and may be applied to any linear imaging system.

Models, Structural↗

Two algorithms for use with an orthogonal-view coded-aperture system.

ohe goal of single photon emission computerized tomography is to map out a three-dimensional distribution of a radionuclide that is concentrated in a structure of interest. There are a number of imaging modalities that achieve this goal with varying degrees of success. In this study, computer simulation is used to explore a novel imaging modality, orthogonal-view coded-aperture imaging. Furthermore, a comparison is made between two reconstruction algorithms, one being an iterative back-projection algorithm and the other a Monte Carlo algorithm. Finally, evidence is presented suggesting that a deficiency in the projection data due to multiplexing is less disturbing than that due to limited angular range.

Evaluation Studies as Topic↗