Contribution of normalization to image noise for the Siemens Inveon small-animal PET scanner

Due to the large number of detector elements (25,600 crystals) and lines of response (LOR) (88,453,120), normalization of the Siemens Inveon small-animal PET scanner can introduce additional image noise when using the standard, individual LOR-based normalization (ILORNORM) of 10 h scan duration. Recently, component-based normalization (CNORM) has been made available for this scanner. The aim of this study was to compare the performance of both normalization algorithms and to determine the minimum time needed for a normalization scan. Normalization scans of 1 min up to 115 h were performed. Reconstructed image noise was determined using the FDG-filled NEMA NU-4 small-animal PET image quality phantom. CNORM significantly reduced image noise as compared to ILORNORM. For CNORM, the scan duration could even be reduced to 10 min without deterioration of reconstructed image quality. Such a short scan duration implies that scanner normalization can easily be incorporated in daily or weekly quality control procedures.