Phantom study
The CRs of four hot spheres decreased with the increase of the penalization factors except HR2.3, and a slight decline of HR3.03 to HR3.5 was observed in three large hot spheres with the diameter of 22 mm (85.4-83.8), 17 mm (77.0-74.4) and 13 mm (78.3-75.2) (Figure 1a). The CRs of HR3.07 to HR3.1were higher than those of O3 for four spheres. The CRs was the highest in the diameter of 22 mm hot sphere, second in 17 mm, third in 13 mm, lowest in 10 mm for HR2.03 to HR2.5, but the CRs in the diameter of 13 mm hot sphere were slightly higher than those attained in the diameter of 17 mm hot sphere for HR3.03 to HR3.5 (Figure 1a). Moreover, the mean of normalized activity was higher than 1.0 for HR2.03 to HR2.07(1.00-1.01), and HR3.03 to HR3.4 (1.01-1.05), the mean of normalized activity was very close to 1.0 for HR2.1 (0.996) and HR3.5 (0.998), and the mean of normalized activity of HR2.2 to HR2.5 (0.92-0.97) was less than 1.0 (Figure 2a).
The BVs decreased along with the increase of the penalization factors for each hot sphere. Moreover, the BVs of O2 were higher than O3 and HYPER iterative 2 m/b and 3m/b groups at the same diameters. The BVs were lowest for 22-mm-diameter sphere, followed by 17 mm and 13 mm diameter spheres, and highest for 10-mm-diameter sphere with the same reconstruction method. The BVs of HR2.03 were higher than those of O3 at the same diameter, and the BVs of HR2.07 and HR2.1 were comparable to O3, because the difference was limited to a small range ( -1.0 % -0.7 %). Furthermore, the BVs of other HYPER iterative groups (HR2.2 to HR2.5 and HR3.03 to HR3.5) had better performance than O3 (Figure 1b).
Quantitative analysis of clinical study
The average liver SUVmean was approximately equal in all reconstruction groups, it was 5.95 for O2, 5.96 for O3 and HR2.03 to HR2.5, and 5.97 for HR3.03 to HR3.5, respectively (Table 2). The CVs of all reconstruction groups had less than 15 %. In detail, the highest CV of liver was 14.36 ± 3.38 % and 12.88 ± 3.26 % for O2 and O3, respectively (Table 2). The CV declined from 14.00 ± 3.00 % to 10.59 ± 3.23 % for HR2.03 to HR2.5, and from 13.52 ± 2.99 % to 10.96 ± 3.11 % for HR3.03 to HR3.5 (Table 2). The HR2.1, HR2.2, HR3.1, and HR3.2 groups were considered noise equivalent groups to O3, because their CVs didn’t make a significant difference (p=0.113, 0.711, 0.079, and 0.287), and the image noise of HR2.03 and HR3.03 was equivalent to O2 (p=0.525 and 0.055). The CVs of HR2.03, HR2.07, HR3.03, and HR3.07 group were significantly higher than that of O3 (all p<0.01), whereas the CVs of HR2.3 to HR2.5 and HR3.3 to HR3.5 were significantly lower than O3 (all p<0.01). Moreover, the mean of normalized CV was range from 0.83 to 1.12 for HR2.03-HR2.5 and 0.85 to 1.06 for HR3.03-HR3.5 (Table 2). The standard deviation of normalized CV was lowest for HR3.03-HR3.1 (all 0.07), and increased with the increased of penalization factor.
The lesion SUVmax decreased with the increase of the penalization factors except for HR2.1 (Table 2). The lesion SUVmax of all HYPER Iterative 2m/b and 3 m/b groups was significantly higher than that of O3 (all p<0.001), and the lesion SUVmax of O2 (10.29 ± 6.14) was comparable to that of O3 (10.28 ± 6.01). The mean of normalized SUVmax for HYPER Iterative groups with 3 m/b was higher than those with 2 m/b when the penalization factor was the same. In detail, the mean of lesions SUVmax increased 22 % - 18 % for HR2.03 to HR2.5, and 25 % -23 % for HR3.03 to HR3.5 compared to O3, respectively (Table 2 and Figure 2b).
The lesions were first divided into small (D<10 mm, n=13, range 7.7-9.9 mm), medium (10 ≤
D<20 mm, n=57, range 10.1-19.7 mm), and large categories (D≥20 mm, n=13, range 20.3-27.3 mm) according to their equivalent diameters. The mean of normalized lesion SUVmax slightly decreased with the increased penalization factor for each category, and the mean of normalized SUVmax for small lesions was higher than those for medium and large lesions in all HYPER iterative 2 m/b and 3 m/b groups except for HR2.4 and HR2.5 at the same penalization factor and acquisition time (Figure 3a). Minor change of average normalized SUVmax was found for large lesions in all HYPER iterative groups, and the minor change was also found for the medium lesions. The mean of normalized SUVmax was higher than 1.0 for large lesions (range 1.06-1.11) and 1.2 for medium lesions (range 1.20-1.23) in all HYPER iterative groups, respectively. Meanwhile, the mean of normalized SUVmax for small lesionswas range from 1.19 to 1.25 in HR2.03-HR2.5 and increased to 1.45-1.47 in HR3.03-HR3.5.
The effect of the injected activity (IA) on the lesion SUVmax of different reconstruction is demonstrated in Figure 3b. The lesions were classified into low (IA< 1.60 MBq/kg, n=35, range 1.01-1.59 MBq/kg) and large injected activity (IA≥1.60 MBq/kg, n=45, range 1.61-2.43 MBq/kg) according to our previous research and PET detection performance [12]. The mean of normalized SUVmax for high injected activity groups was slightly higher than that for low injected activity in all HYPER iterative groups except for HR2.3-H2.5 (Figure 3b). Detailly, the mean of normalized lesion SUVmax for all HYPER iterative groups floated around 1.2, the range was 1.19-1.23 for low activity and 1.18-1.27 for high activity. The mean of normalized SUVmax of HR3.03 to HR3.5 was comparable and a little higher than that of HR2.03 to HR2.5 for low and large injected activity, respectively.
To investigate the effect of patient weight on the lesion SUVmax, the patients were further sorted into two categories by body mass index (BMI): under-and-normal weight group (BMI <24 kg/m2, n=47, range 18.36-23.72 kg/m2) and overweight group (BMI≥24 kg/m2, n=33, range 24.22-28.12 kg/m2). The mean of normalized lesion SUVmax for under-and-normal weight group was higher than that for overweight group in the HYPER iterative 2m/b groups at the same penalization factor, while the trend was contrary with 3m/b acquisition. Additionally, the mean of normalized lesion SUVmax had small changes between two BMI-categories in all HYPER iterative groups with the same penalization factor and acquisition time, but obvious increase was found in the HYPER iterative groups with 3 m/b acquisition compared to 2 m/b (Figure 3c). The mean of normalized lesion SUVmax was higher than 1.2 (range 1.20-1.23) and 1.15 (range 1.15-1.27) for under-and-normal weight group and overweight group with HYPER Iterative, respectively.
A total of 80 68Ga-DOTANOC avid lesions were identified: Eight lesions in the lung, 38 in the bone, two in the thyroid, 27 in the lymph node, four in the soft issue, and one in the pancreas. The mean of normalized SUVmax was weakly decreased with the increased penalization factor for lung, lymph node and bone metastases. Moreover, the mean of normalized SUVmax for HYPER Iterative groups with 3 m/b acquisition was higher than that with 2 m/b when the lesion location and the penalization factor were the same. Of note, the mean of normalized lesion SUVmax was highest in bone, second in lymph node, lowest in lung at the same penalization factor. Furthermore, the mean of normalized lesion SUVmax was higher than 1.0 (range 1.07-1.18) for lung, 1.1 (range 1.11-1.21) for lymph node, and 1.2 (range 1.26-1.30) for bone with HYPER Iterative, respectively. (Figure 3d).
Qualitative image quality of clinical study
The mean of image quality score was first increased then declined with the increased of the penalization factors (Figure 4). The highest image quality score was assigned to HR2.2 (3.20 ± 0.52) and HR3.1 (3.70 ± 0.36) for 2 m/b and 3 m/b acquisition groups. The lowest score was acquired at HR2.4 (2.44 ± 0.45) and HR2.5 (2.16 ± 0.35) due to poor contrast for small lesions (Figure 7 and Figure 9), and the secondary score was given to O2 (2.81± 0.35) because of suboptimal image noise (Figure 5, Figure 6, and Figure 8). The average scores of HR3.07 to HR3.3 were significantly high than that of O3 (all p<0.05), and the image quality scores of HR2.07 to HR2.3 did not differ from O3 (all p>0.062). The inter-rater agreement was substantial (k=0.71).