Statistical normalization of 3D panel PET detector

A statistical normalization technique was developed for the CPS dual-layer HRRT scanner. A component based model was proposed to include crystal sensitivity, geometric response, and layer identification factors. A maximum likelihood based approach was used to estimate these factors. Emission data were acquired on the HRRT scanner with a rotating rod source in listmode format to preserve counts information in each line of response. A conjugate gradient algorithm was implemented to obtain the maximum likelihood estimation of normalization factors. These factors were then used to construct the normalization sinogram array. Uniform phantom data were acquired and reconstructed to compare the direct to the new normalization technique. The results show that the new normalization technique achieves better image noise levels than the direct technique. The improved technique could be used to shorten the acquisition time of the normalization