Abstract
1748
Objectives: It was recently suggested that PET images generated from the sum of multiple short duration frames have superior SNR than those generated from a single acquisition over the same scan duration (JNM vol 47(5) 374P 2006, IEEE medical imaging conf. M04-5 2006). In this abstract we focus on the variation of SNR with respect to the number of frames used to generate the final PET image. Methods: A NEMA IEC phantom containing six spheres was scanned on a DISCOVERY STE PET/CT scanner. The phantom was scanned in 2D and 3D using List mode in a single filed of view (FOV). Three sphere to background ratios (SBR) were used (3.5:1, 5.34:1, 8:1) while keeping background activity concentration constant. In each case, the total scan duration was 3 minutes (similar to our clinical protocol/FOV). The List data was then segmented into 14 combinations of short-duration frames: 180*1, 90*2, 60*3, 36*5, 30*6, 20*9, 18*10, 12*15, 9*20, 6*30, 5*36, 3*60, 2*90, and 1*180 (frames*seconds). For each combination, the frames were reconstructed using OSEM (2 iterations 21 subsets), and summed to produce the 3-minute image. ROIs were drawn on all 6 spheres of the resulting images and the mean value was recorded. The standard deviation (STD) of 30 to 60 random pixels chosen from the background region was used to evaluate the noise in the background. The SNR for each sphere, acquisition mode, and SBR was then calculated by dividing the mean by the STD. The SNR was then plotted for the different frame combinations in each SBR. Results: Comparisons within each case revealed that as the number of frames decreased, the mean ROI intensity of all spheres increased slowly before reaching a plateau, while the noise first decreased and then increased. This resulted in a maximum SNR that ranged between 12*15 and 6*30 for 2D and 30*6 and 18*10 (frames*seconds) for 3D. Similar observations were found for different SBRs. In all cases, images generated from the combination with the largest number of frames had the worst SNR and were of lower quality than images generated for a single frame. Conclusions: This study showed that images generated from multiple frames would have better quality only if the proper combinations of frames and durations were used.
- Society of Nuclear Medicine, Inc.