A study of count-rate dependent normalization effect in 3D-PET data

Accurate normalization correction of the efficiency of each line-of-response (LOR) is essential in the quantitative PET reconstruction. In this paper, we first studied the effect of count-rate dependent normalization in 3D PET reconstructed images in a recent developed PET prototype. The results show that, with mis-matched normalization correction, the reconstructed images show severe "ring" artifact. A count-rate dependent normalization calibration method is proposed to minimize the artifacts which are caused by the inaccurate normalization coefficients (NC) calibrated at a different system count-rate. In this work, the fully-3D extended Defrise algorithm was used for the calculation of the count-rate dependent detector block profile NCs. A small F18-fillable cylinder phantom located at the center of the scanner field of view (FOV) was used for data calibration.