The improvement of SPECT images using scatter correction techniques
The removal of scattered radiation is recognised as one of the major goals to be achieved in SPECT. In this work three scatter removal techniques were investigated. These were dual window scatter subtraction, deconvolution using a Wiener filter and energy weighted acquisition (EWA) using a weighted acquisition module (WAM). In addition to this, the effects of simple background subtraction on SPECT images was also investigated. The techniques were investigated in both a semi-quantitative manner, in terms of the effects of each technique on the cold lesion contrast and image mottle, and qualitatively, in terms of the ability of observers to perform a detection task. This was done using Relative Operating Characteristic (ROC) experiments. Each technique is investigated individually to discover the parameters needed to optimise performance. The results showed that the optimum parameters for each scatter reduction technique was dependent on the measure of image quality used for optimisation and showed that all of the techniques investigated produced better results than those produced using the standard 20&'37 photopeak acquisition approach, however, in general no significant difference could be found between the techniques. The image noise produced by each technique was also evaluated by calculating the noise power spectra (NPS) produced by each technique. The calculated NPS showed the spatial content of the noise produced by each technique was different for each scatter reduction technique. The results in this work showed that empirically defined measures of image quality are poor predictors of observer performance.