Page 122 - 南京医科大学自然版

第45卷第11期
·1652 · 南 京 医 科 大 学 学 报 2025年11月

A B 2.4 C
25%~75%
16 1.8
1.5IQR 2.0
Median
12 1.5
Mean 1.6
R SUV mex 8 Outliers R SUV mean 1.2 R SUV mean 1.2
0.8
4 0.9
0.4
0 0.6
0
10% 30% 50% 10% 30% 50% 10% 30% 50%
The proportion of activity used The proportion of activity used The proportion of activity used
to the current clinical activity to the current clinical activity to the current clinical activity
18 18
of all voxels in patient PET images with SUV values
of PET images. B：Effect of F⁃FDGactivity on R SUV mean
A：Effect of F⁃FDG activity on R SUV max
18
of all voxels in patient PET images with SUV values greater than 1.
greater than 0.1. C：Effect of F⁃FDG activity on R SUV mean
图1 18 F⁃FDG活度对PET影像SUV值的影响
18
Figure 1 Effect of F⁃FDG activity on SUV values in PET images
可以看出，采用相同的 PET 影像重建算法，当所 2.3 基于深度学习的低 F⁃FDG给药活度下的PET
18
用 F⁃FDG 的 活 度 降 低 时 ，患 者 肿 瘤 病 灶 内 的 影像质量恢复
18
18
SUVmean和SUVmax均呈现增大趋势，其中SUVmax 基于收集到的不同 F⁃FDG 给药活度下的患者
的变化更显著。当 F⁃FDG的注射活度降低到临床 PET/CT 影像数据，本研究搭建了 SwinUNetR⁃GAN
18
现行活度的10%时，肿瘤病灶的SUVmean增大了约 网络以提高低 F⁃FDG 给药活度下患者PET 影像质
18
1.1倍，而SUVmax增大了约2倍。 量，首先采用视觉分析法对SwinUNetR⁃GAN网络的

A B
2.0 30
25
1.6
20
SUVmean 0.8 SUVmax 15
1.2
10
0.4
5
0
0
10% 30% 50% 100% 10% 30% 50% 100%
The proportion of activity used The proportion of activity used
to the current clinical activity to the current clinical activity
C 1.6 D 10
8
1.4
R SUV mean 1.2 R SUV max 6

4
1.0
2

0.8 0
10% 30% 50% 10% 30% 50%
The proportion of activity used The proportion of activity used
to the current clinical activity to the current clinical activity
18
18
A：Effect of F⁃FDG activity on SUVmean of tumor. B：Effect of F⁃FDG activity on SUVmax of tumor. C：Ratio of SUVmean of tumor at different
18 18
F⁃FDG activities to SUVmean at current clinical activity used. D：Ratio of SUVmax of tumor at different F⁃FDG activities to SUVmax at current clinical
activity used.
图2 18 F⁃FDG活度对肿瘤SUV值的影响
18
Figure 2 Effect of F⁃FDGactivity on tumor SUV value