Preserving Poisson characteristics of PET data with weighted OSEM reconstruction

To evaluate the importance of preserving the Poisson characteristics of PET data when using OSEM reconstruction, the authors have examined several data weighting schemes with increasing complexity which progressively account for attenuation, normalization, random coincidences and Compton scatter with both ordinary and shifted Poisson models. All schemes have first been tested on 2D dynamic data from a decaying phantom filled with H₂¹⁵O. The images produced by the various weighting schemes were compared to the ones obtained by filtered backprojection reconstruction. Small positive bias (1%) was detected for the FW and SP schemes and may be due to implementation. The impact of these reconstruction methods on physiological parameters is illustrated with clinical protocols measuring myocardial blood flow with ¹³NH₃ and Standard Uptake Values with ¹⁸FDG. Despite the high random rate in cardiac studies, no positive bias was detected in the images when using the AW and ANW schemes with scans precorrected for random and scatter. For whole-body studies, at low random rates, the ANW scheme produced the same quantitative results as the FW and SP ones. That scheme is likely the best choice for obtaining high quality emission images and unbiased ROI´s in most situations on current 2D scanners