SPECT is a medical diagnostic procedure. The patient receives some type of radiation isotope which emits gamma photons, so the spatial distribution can be measured. For the detection of photons gamma camera is used. The parts of the camera: a collimator, scintillation crystal, photomultiplier tubes, and electronics.
The traditional approach to tomographic image reconstruction is filtered back projection based on Radon transform, but lately iterative reconstruction techniques getting common in practice.
Compared to the ideal case, various factors can reduce the quality of the resulting images. These include the gamma photons and matter interactions, such as the Compton scattering and attenuation.
Compton scattered photons are detected insufficient and thus reduces the quality of the reconstructed images so corrections are necessary. Several possible solutions, such as the use of two energy window or solutions based on analytical equations, have been developed.
I’m working on a scatter correction algorithm which will be installed in an iterative image reconstruction process. As a first step, I wrote a sampling algorithm for the possible gamma photon source-points and scattering-points. The third version of the algorithm samples the points in proportion to the density values of space, in accordance with the criteria Poisson-Disk. Based on the processed literature, the generated sample shows Poisson distribution. The current algorithm can handle a maximum of 500,000 points and is suitable for sampling up to 10243 size images.
After the sampling algorithm was made, I wrote the Compton scatter estimation algorithm. The process properly followed the path of gamma photon’s life, during the attenuation and scattering phenomena, from the source point to the detector. The algorithm takes into account the attenuation, the scattering probability, the reduction of photon energy because of scattering, associated with the energy windowing, and the point-spread-function of the detector as well. It is possible now to take account the secondary scatter as well as the first scatter. After that I built it into an iterative process and as a last step I tasted the algorithm.