2.1Visual evaluation for the image quality
In the full-dose group, OSEM-10s, OSEM-30s and OSEM-1min images were noisier and characterized by roughness and poor homogeneous (Fig.1), which were scored 1.00±0.00, 1.90±0.32 and 2.80±0.42, respectively. These three groups could not meet the high quality requirement for clinical diagnosis (Table 2). The image quality of OSEM-2min reached the nearly excellent level with a score of 4.80±0.84(Fig.1and Table 2). As increased the scan time, the images of the OSEM-3min to OSEM-5min were all excellent (Scores, 5.00±0.00) (Fig.1and Table 2). For the HYPER Iterative reconstructed PET images, HYPER Iterative-10s and HYPER Iterative-30s had smaller noise and were better than their counterparts using OSEM(Fig.1). The corresponding scores were 2.80±0.42 and 4.10±0.32, respectively, which were significantly higher than those of OSEM-10s and OSEM-30s(P<0.003) (Table 2), but still did not reach the excellent level. The image quality of PET nearly reached the excellent level at 1min with a score of 4.90±0.32(Fig.1and Table 2). After that, images of HYPER Iterative-2min to HYPER Iterative-5min were all excellent with each score of 5.00±0.00(Fig.1 and Table 2).
In half-dose group, OSEM-10s, OSEM-30s, OSEM-1min and OSEM-2min were worse than their counterparts of the full-dose group (Fig.2), and their scores were 1.00±0.00, 1.80±0.42, 2.70±0.48, and 3.80±0.42, respectively (Table 2). The quality of OSEM-3min images nearly reached the excellent level with a score of 4.90±0.32. Both OSEM-4min and OSEM-5min images reached excellent with a score of 5.00±0.00. The images of HYPER Iterative-10s, HYPER Iterative-30s and HYPER Iterative-1min were better in quality than those of their counterparts reconstructed using OSEM. Their scores were 2.00±0.00, 3.00±0.00 and 3.80±0.42, respectively, but not reached an excellent level. The image quality of PET reconstructed by HYPER Iterative reached the excellent level at 2 min with a score of 4.80±0.42. The images from HYPER Iterative-3min to HYPER Iterative-5min were all excellent (Score, 5.00±0.00).
2.2 Quantitative evaluation of PET/CT image quality
2.2.1 The effect of two reconstruction algorithms on homogeneous radioactivity areas
In both the full-dose and half-dose groups, the shorter image acquisition time resulted in the larger SD in the liver and mediastinal blood pool. Compared to the HYPER Iterative reconstruction, OSEM reconstruction had a higher SD in each scan time from 10 sec to 5 min (all P<0.05), but the difference turned smaller as increasing the acquisition time (Table 3). For the full-dose group, the SD by OSEM reconstruction with scan time of 10 sec was up to 4~5 times higher than that by the HYPER Iterative reconstruction(Table 3).The difference increased to about 8 times higher in the liver in the half-dose group at 10 sec (Table 3). The SDs generated by OSEM reconstruction or HYPER Iterative were found to be larger in half-dose than that in full-dose.
The SUVmax of blood pool and liver reconstructed by OSEM was significantly higher than that by HYPER Iterative reconstruction at 10 sec to 2 min. The falsely higher SUVmax in blood pool and liver homogeneous radioactivity areas reconstructed by OSEM occurred within 2 min, with up to about 28%~34% higher in full dose at 10 sec and up to 42%~50% higher in half-dose at 10 sec(Table 4).With the scan time was extended to 2-5 min, the SUVmax of blood pool and the liver gradually decreased to a stable small level. Compared to the full-dose, the falsely higher SUVmax in blood pool reconstructed by OSEM was more obvious in half-dose (Table 4).
2.2.2 The effect of two reconstruction algorithms on large and small positive lesions
For large positive lesions≥2.0cm, the SUVmax reconstructed by HYPER Iterative in the full-dose was significantly higher than that reconstructed by OSEM in 2 min to 5 min(P<0.05), which was about 9% higher at 5 min(Fig.3,5 and Table 5). A similar trend occurred for the half-dose with up to 13% higher at 5 min (Fig.4,5 and Table 5). However, no significant difference of SUVmax between two reconstruction algorithms was observed within 1 min(P>0.05), either in the full-dose or the half-dose. For small positive lesions≤10mm, the SUVmax of HYPER Iterative reconstruction in the full-dose was higher at 1 min to 5 min(P<0.05) than that of OSEM reconstruction, which was 45% higher at 5 min. In the half-dose, this difference was increased up to 75% at 5 min(P<0.05). No significantly difference of SUVmax in small lesions was found between two reconstruction algorithms within 30 sec(P>0.05), either in the full-dose or the half-dose.
Similar to SUVmax, the TBR of large lesions reconstructed by HYPER Iterative was higher than that reconstructed by OSEM within 2 min to 5 min(P<0.05). HYPER Iterative reconstruction was 9% higher than OSEM reconstruction in the full-dose at 5 min and 23% higher than OSEM reconstruction in the half-dose at 5 min (Fig. 6) (Table 6). However, no significant difference between two reconstruction algorithms was observed within 1 min(P>0.05). For small lesions, the TBR of small lesions reconstructed by HYPER Iterative was higher than that reconstructed by OSEM within 1 min to 5 min(P<0.05). At 5 min, HYPER Iterative reconstruction was 45% and 94% higher in the full-dose and half-dose groups, respectively (Fig.6 and Table 6). Although no significantly difference was found between two reconstruction algorithms within 30 sec(P>0.05).