Clinical application of maximum entropy image processing in planar radionuclide imaging

The aim of the research was to develop the technique of maximum entropy for clinical application and to evaluate its effectiveness in improving image quality.Maximum entropy (ME) requires definition of various parameters for its operation. Studies were carried out to investigate the way in which the values of these parameters affected image quality. This allows methods for deriving optimal values in any situation to be devised.A study was carried out to investigate the way in which the figure of merit (FOM) method of image quality assessment depended on definition of regions defining the object and surroundings. This allowed description of a method for defining the regions which provided robust values for FOM.A comparative evaluation of maximum entropy processing with simple image smoothing (i.e. conventional smoothing (SM)) and Wiener filtering (WF) was carried out in simulated images of a planar object. Image quality was evaluated using the FOM and using receiver operating characteristic (ROC) analysis with two different observers. The FOM analysis showed that all image processing technique produced significant improvement over the raw data and that ME was the best of the methods. These findings were generally supported by the ROC analysis although, the conclusions were not so clearly defined. There was significant correlation between FOM and detectability for individual observers interpreting images from a single processing technique. Correlation was poorer when data from all the methods were combined.A further comparative evaluation of the processing techniques in simulated lung images was performed using ROC analysis. The analysis failed to show significant improvements in detectability using conventional smoothing or Wiener filtering.