Evaluation of a self-normalization method for two PET cameras using prompt, delay or total emission data

For PET cameras the non-uniformity of the sensitivity of the lines of response (LORs) has to be corrected with appropriate normalization coefficients (NCs) in order to generate artifact-free quantitative images. The NCs are commonly estimated by dedicated calibration scan in which the scanner performance is measured for a known activity source. With the conventional normalization methods, proper handling of rate dependency of the NCs is not an easy task. To have the best results, the scanning conditions for the normalization scan and object scan should be similar. So, it would be ideal to calculate the NCs from the emission data itself (self-normalization). The self-normalization method would also eliminate the need for extra normalization scans. In this work, we report on evaluation of a new self-normalization method for two high-resolution cameras: a small animal PET camera and a whole-body PET camera. Several phantoms and mice were scanned. For calculation of crystals efficiencies, for each scan, we used three sets of data and compared them: prompt window data, delay window data and sum of the data in both windows (total). The tested self-normalization method generated artifacts free images. For both cameras, we did not observe any significant differences between the images obtained from data corrected with NCs generated from only prompt data, or random data, or all the data. Currently, we are using all the emission data, for better statistics, in calculations of NCs.