Maximum entropy image restoration in nuclear medicine
Maximum entropy image restoration in nuclear medicine
A model for the process involved in radioisotope emission imaging is a distribution of Poisson emission processes whose rates, which depend on the intensity of radioisotope concentration, are to be estimated on the basis of counting records from an array of detectors. Since emission imaging is particularly prone to problems arising from the limited resolution of the detectors, combined with weak statistics of the same, the resulting images require substantial processing. In this thesis the problem of planar image restoration from noisy measurements as encountered in Nuclear Medicine is considered. A new approach for treating the measurements prior to restoration is presented where they are represented by a spatial noncausal interaction model. The measurements were found to be highly correlated from independent measurements and tests performed on actual gamma camera records. A simultaneous autoregressive model is utilized to describe these correlations. In conventional nuclear medicine imaging this correlation results in non-uniformity in gain across the detector array. The restoration problem is treated here as a constrained optimisation approach where the Maximum Entropy principle is used to select the image with minimum information content consistent with the available measurements. Regarding the constraint equations, which are specified in terms of the available measurements and the dispersive characteristics of the detection process, a model is introduced. This is based on the Poisson statistics of the emissions and a point spread function for the imaging system. Although the algorithms treated here are iterative and hence slower than direct techniques, of the kind favoured in clinical practice, they are aimed at producing further improvement in image quality. In developing the algorithms, consideration has been given to the need to compromise between modelling the complex stochastic nature of the emission and focusing processes, and to reduce the computational cost, particularly in time, for producing a complex image whereresolution and sensitivity are paramount. Results for both simulated data and actual gamma camera data are presented and compared with more conventional techniques. The latter include pelvic and heart cycle imaging results.
University of Southampton
1989
Oliveira, Vilma Alves de
(1989)
Maximum entropy image restoration in nuclear medicine.
University of Southampton, Doctoral Thesis.
Record type:
Thesis
(Doctoral)
Abstract
A model for the process involved in radioisotope emission imaging is a distribution of Poisson emission processes whose rates, which depend on the intensity of radioisotope concentration, are to be estimated on the basis of counting records from an array of detectors. Since emission imaging is particularly prone to problems arising from the limited resolution of the detectors, combined with weak statistics of the same, the resulting images require substantial processing. In this thesis the problem of planar image restoration from noisy measurements as encountered in Nuclear Medicine is considered. A new approach for treating the measurements prior to restoration is presented where they are represented by a spatial noncausal interaction model. The measurements were found to be highly correlated from independent measurements and tests performed on actual gamma camera records. A simultaneous autoregressive model is utilized to describe these correlations. In conventional nuclear medicine imaging this correlation results in non-uniformity in gain across the detector array. The restoration problem is treated here as a constrained optimisation approach where the Maximum Entropy principle is used to select the image with minimum information content consistent with the available measurements. Regarding the constraint equations, which are specified in terms of the available measurements and the dispersive characteristics of the detection process, a model is introduced. This is based on the Poisson statistics of the emissions and a point spread function for the imaging system. Although the algorithms treated here are iterative and hence slower than direct techniques, of the kind favoured in clinical practice, they are aimed at producing further improvement in image quality. In developing the algorithms, consideration has been given to the need to compromise between modelling the complex stochastic nature of the emission and focusing processes, and to reduce the computational cost, particularly in time, for producing a complex image whereresolution and sensitivity are paramount. Results for both simulated data and actual gamma camera data are presented and compared with more conventional techniques. The latter include pelvic and heart cycle imaging results.
This record has no associated files available for download.
More information
Published date: 1989
Identifiers
Local EPrints ID: 461190
URI: http://eprints.soton.ac.uk/id/eprint/461190
PURE UUID: f5c400d5-05b7-4da7-89fa-07d96488deed
Catalogue record
Date deposited: 04 Jul 2022 18:38
Last modified: 04 Jul 2022 18:38
Export record
Contributors
Author:
Vilma Alves de Oliveira
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics