The application of Monte Carlo simulation to the design of collimators for single photon emission computed tomography.

摘要

Single photon emission computed tomography offers the potential for quantification of the uptake of radiopharmaceuticals in-vivo. This thesis investigates some of the factors which limit the accuracy of these methods for measurements in the human brain and investigates how the errors can be reduced. Modifications to data collection devices rather than image reconstruction techniques are studied. To assess the impact of errors on images, a set of computer generated test objects were developed. These included standard Anger and Phelps phantoms and a series of slices of the human brain taken from an atlas of transmission tomography. System design involves a balance between resolution and noise in the image. The optimal resolution depends on the data collection system, the uptake characteristics of the radiopharmaceutical and object size. A method to determine this resolution was developed and showed a single-slice system employing focused, probe detectors to offer greater potential for quantification in the brain than systems based on multiple Anger gamma cameras. A collimation system must be designed to achieve the required resolution. Classical, geometric design is not satisfactory in the presence of scattering materials. For this reason a Monte Carlo simulation allowing flexible choice of collimator parameters and source distribution was developed. The simulation was fully tested and then used to predict the performance of collimators for probe and camera based systems. These assessments were carried out for the 'worst case source' which was a concept developed and validated to allow faster prediction of collimator performance. In essence the geometry of this source is such as to allow a resolution measurement to be made which represents the worst value expected from the system. The effect of changes in collimation on image quality was assessed using the computer phantoms and simulation of the data acquisition process on the singleslice system. These data were reconstructed with proprietary software. Analysis of these images showed that improved collimator resolution facilitated similarly improved image resolution. A novel method of determining resolution from the effect of partial volume on a Phelps phantom was developed for these measurements.