摘要
目的探讨基于深度学习(DL)重建列表模式提取低剂量图像用于儿童^(18)F-FDG PET/CT检查的可行性。方法对93例患儿行全身^(18)F-FDG PET/CT,分别以120秒/床位行标准全剂量扫描、20秒/床位行快速扫描,之后以列表模式于120秒/床位图像中提取20秒/床位图像。将20秒/床位图像与以列表模式提取的20秒/床位图像传输至DL平台进行重建,分别获得相应全剂量图像(DL-f20s、DL-20s);对比标准全剂量120s、DL-f20s、DL-20s图像的靶本比(TBR)、对比噪声比(CNR)、变异系数(CV)、图像质量主观评分,以及肝血池、纵隔血池、肌肉血池及原发病灶最大标准摄取值(SUV max)和平均标准摄取值(SUV mean),分析基于DL-f20s与DL-20s图像所获SUV max及SUV mean的一致性。结果120s、DL-f20s和DL-20s图像的主观质量Likert评分、TBR、CNR、CV及各血池、病灶的SUV max、SUV mean差异均无统计学意义(P均>0.05)。DL-f20s与DL-20s图像之间,5%以上肌肉血池SUV max、SUV mean差值在95%一致性界限(LoA)以外,其余血池及病灶的SUV max、SUV mean差值在95%LoA以外者均≤5%。结论基于DL重建列表模式提取低剂量图像可用于儿童^(18)F-FDG PET/CT检查,且所获图像的质量及代谢参数与快速扫描相当。
Objective To investigate the feasibility of extracting low-dose images of^(18)F-FDG PET/CT using list mode based on deep learning(DL)reconstruction in children.Methods Whole body^(18)F-FDG PET/CT was performed in 93 pediatric patients using 120 s/bed standard full-dose scanning and 20 s/bed fast scanning,and the 20 s/bed images were extracted from 120 s/bed images using list mode.Then 20 s/bed images and 20 s/bed images extracted using list mode were transferred to the DL platform for reconstruction,and the relative corresponding full-dose images(DL-f20s,DL-20s)were acquired,respectively.The target to background ratio(TBR),contrast to noise ratio(CNR),coefficient of variation(CV),subjective score of imaging quality,the maximum standard uptake value(SUV max)and mean standard uptake value(SUV mean)of the liver blood pool,mediastinal blood pool,muscle blood pool and primary lesion were compared among standard full-dose 120s,DL-f20s and DL-20s images,and the consistencies of SUV max and SUV mean obtained from DL-f20s and DL-20s were analyzed images.Results There was no significant difference of imaging subjective quality Likert score,TBR,CNR,CV,nor of SUV max and SUV mean of each blood pool and primary lesions among 120s,DL-f20s and DL-20s images(all P>0.05).Between DL-f20s and DL-20s images,more than 5%differences of SUV max and SUV mean of muscle blood pool were outside the 95%limits of agreement(LoA),while≤5%differences of SUV max and SUV mean of the other blood pools and lesions were outside the 95%LoA.Conclusion Low-dose images extracted using list mode based on DL reconstruction could be applied in^(18)F-FDG PET/CT examination in children,which had comparable imaging quality and metabolic parameters with those images obtained from fast scanning images.
作者
马欢
冯莉娟
鲁霞
张抒欣
阚英
李楠
张辉
杨吉刚
MA Huan;FENG Lijuan;LU Xia;ZHANG Shuxin;KAN Ying;LI Nan;ZHANG Hui;YANG Jigang(Department of Nuclear Medicine,Beijing Friendship Hospital,Capital Medical University,Beijing 100050,China;SinoUnion Healthcare Inc.,Beijing 100192,China;Department of Biomedical Engineering,School of Medicine,Tsinghua University,Beijing 100084,China)
出处
《中国介入影像与治疗学》
北大核心
2023年第4期233-237,共5页
Chinese Journal of Interventional Imaging and Therapy
基金
国家自然科学基金项目(81971642).
