Page 119 - 南京医科大学自然版

第45卷第11期 南京医科大学学报（自然科学版）
2025年11月 Journal of Nanjing Medical University（Natural Sciences） ·1649 ·

·临床研究·

基于Swin Transformer的低活度PET影像质量恢复方法研究

丁 威，唐立钧，田 锋 *
南京医科大学第一附属医院核医学科，江苏 南京 210029

［摘 要］ 目的：探究氟代脱氧葡萄糖（fludeoxyglucose，F⁃FDG）（活度对正电子发射计算机断层成像（positron emission tomog⁃
18
raphy，PET）CT影像质量的影响规律，建立PET影像质量恢复方法实现对患者的辐射防护。方法：基于 Swin Transformer 全局
特征识别低活度 PET 影像质量恢复深度学习网络 SwinUNetR⁃GAN，随机选择于 2024 年行 F⁃FDG PET/CT 检查的124例
18
患者影像，探究不同给药活度下PET 影像质量下降规律，并基于SwinUNetR⁃GAN 网络实现低活度 PET 影像质量的恢复。结
果：随着 F⁃FDG活度的降低，患者PET影像中正常组织及肿瘤病灶内的标准摄取值（standard uptake value，SUV）均值（SUVmean）
18
和最大值（SUVmax）均呈现增大趋势，当将 F⁃FDG的注射活度降低到临床现行活度的10%时，肿瘤病灶的SUVmean增大了约
18
1.1倍，而SUVmax增大了约2.0倍。此外，采用SwinUNetR⁃GAN网络可降低10%活度PET影像的噪声，绝对偏差由0.21降低至
0.15，相对偏差可由0.33降低至约0.25。结论：明确了 F⁃FDG活度对患者PET影像中正常组织及肿瘤组织定量参数的变化规
18
18
律，提出了专用于低 F⁃FDG 活度PET影像质量恢复的SwinUNetR⁃GAN 网络，实现了降低患者所受辐射剂量的同时确保PET
影像的疾病诊断性能。
［关键词］ 18 F⁃FDG；活度；辐射剂量；Swin Transformer；深度学习
［中图分类号］ R814.42 ［文献标志码］ A ［文章编号］ 1007⁃4368（2025）11⁃1649⁃07
doi：10.7655/NYDXBNSN250838

Research on low⁃activity PET image quality restoration method based on Swin Transformer

DING Wei，TANG Lijun，TIAN Feng *
Department of Nuclear Medicine，the First Affiliated Hospital of Nanjing Medical University，Nanjing 210029，China

18
［Abstract］ Objective：To explore the influence of F⁃FDG activity on PET/CT image quality and establish a PET image quality
restoration method to achieve radiation protection for patients. Methods：A deep learning network SwinUNetR⁃GAN for restoring the
quality of low⁃activity PET images based on the global feature recognition of Swin Transformer was established. The images of 124
18
patients who underwent F ⁃ FDG PET/CT examinations in 2024 were randomly selected to explore the law of PET image quality
18
degradation under different activities of F⁃FDG，and the quality of low⁃activity PET images was achieved based on the SwinUNetR⁃
GAN network. Results：As the activity of F⁃FDG decreases，the mean and maximum values of the standard uptake value（SUV）
18
18
（SUVmean and SUVmax）in normal tissues and tumor lesions in the PET images show an increasing trend. When the activity of F⁃
FDG is reduced to 10% of the current clinical activity，the SUVmean of the tumor lesion increases by about 1.1 times，while the
SUVmax increases by about 2.0 times. In addition，the use of the SwinUNetR⁃GAN network can reduce the noise of 10% activity PET
images，The absolute deviation decreased from 0.21 to 0.15，and the relative deviation could be reduced from 0.33 to about 0.25.
Conclusion：This study clarifies the change law of quantitative parameters of normal tissues and tumor tissues in patient PET images
18
18
under activity of F⁃FDG，and then，SwinUNetR⁃GAN network dedicated to low F⁃FDG activity PET image quality restoration is
proposed，which could achieve the disease diagnosis performance of PET images while reducing the radiation dose deposited in patient.
［Key words］ 18 F⁃FDG；activity；radiation dose；Swin Transformer；deep learning
［J Nanjing Med Univ，2025，45（11）：1649⁃1655］

［基金项目］ 国家自然科学基金青年基金（12405387）；江苏省基础研究计划自然科学基金青年基金（BK20241107）；江苏省卫
生健康委科研课题（Ym2023101）
通信作者（Corresponding author），E⁃mail：tf_0145@163.com（ORCID：0009⁃0009⁃6277⁃1402）
∗