| 田燕,龙其刚,许振春,张文倩,蔡亮.深度渐进重建算法用于重建全身18F-FDG PET图像[J].中国医学影像技术,2025,41(1):142~147 |
| 深度渐进重建算法用于重建全身18F-FDG PET图像 |
| Deep progressive reconstruction algorithm applicated in reconstructing whole-body 18F-FDG PET images |
| 投稿时间：2024-07-27 修订日期：2024-10-18 |
| DOI：10.13929/j.issn.1003-3289.2025.01.030 |
| 中文关键词: 正电子发射断层显像术 人工智能 图像质量 |
| 英文关键词:positron-emission tomography artificial intelligence image quality |
| 基金项目: |
|
| 摘要点击次数: 386 |
| 全文下载次数: 113 |
| 中文摘要: |
| 目的 观察深度渐进重建(DPR)算法用于重建全身18F-FDG PET图像的价值。方法 回顾性纳入67例接受全身18F-FDG PET/CT显像患者,分别利用有序子集最大期望值法(OSEM)及DPR算法重建设备列表下每床位30、60、90及120 s 数据,共得到7组PET图像,包括3组OSEM(OSEM_30、OSEM_60及OSEM_120组)和4组DPR(DPR_30、DPR_60、DPR_90及DPR_120组)图像;比较各组图像主、客观评价结果,后者包括病灶和肝脏最大及平均标准摄取值(SUV),即SUVmax及SUVmean;计算靶区本底比值(TBR)、信噪比(SNR)、对比度噪声比(CNR)及肝脏变异系数(CVliver)。以OSEM_120组为参考,绘制Bland-Altman图,观察基于DPR_30、DPR_60及DPR_90组所获病灶及肝脏SUV与基于OSEM_120组所获结果的一致性。结果 相同采集时间下,各DPR组图像主观评分均高于相应各OSEM组;基于各DPR组所获病灶SUVmax、SUVmean,TBR、SNR、CNR及CVliver均优于相应OSEM组(P均＜0.001)。相比OSEM_120组,DPR_30组主观评分及SNR均降低而TBR及CVliver升高,DPR_60组及DPR_90组图像主、客观评价结果改善(P均＜0.05)或差异无统计学意义(P均＞0.05)。基于上述7组所测肝脏SUVmean差异无统计学意义(P=0.955),且基于DPR_30、DPR_60及DPR_90组所测病灶和肝脏SUVmax及SUVmean均与基于OSEM_120组测量结果具有良好一致性。结论 DPR算法用于重建全身18F-FDG PET图像可在保证图像质量的前提下缩短采集时间。 |
| 英文摘要: |
| Objective To observe the value of deep progressive reconstruction (DPR) algorithm for reconstructing whole-body 18F-FDG PET images. Methods Totally 67 patients who underwent whole-body 18F-FDG PET/CT were retrospectively enrolled. PET data of 30 s, 60 s, 90 s and 120 s per bed in equipment list were reconstructed using ordered subset expectation maximization (OSEM) and DPR algorithms, respectively. Finally 7 groups of reconstructed images were obtained, including OSEM_30, OSEM_60 and OSEM_120, also DPR_30, DPR_60, DPR_90 and DPR_120 groups. The subjective scores, also objective evaluation indexes, i.e. the maximum and mean standard uptake values (SUV) of lesions and livers, namely SUVmax and SUVmean, were compared, and target-to-background ratio (TBR), signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR) and coefficient of liver variation (CVliver) were calculated. Taken results based in OSEM_120 group as references, Bland-Altman plot was drawn to explore the consistency of SUV of lesions and livers obtained based on DPR_30, DPR_60 and DPR_90 groups with those in OSEM_120 group. Results Under the same acquisition time, subjective scores, SUVmax and SUVmean of lesions, TBR, SNR, CNR and CVliver in DPR_30, DPR_60 and DPR_120 groups were superior to those in corresponding OSEM_30, OSEM_60 and OSEM_120 groups (all P＜0.001). Compared with OSEM_120 group, subjective scores and SNR decreased but TBR and CVliver increased in DPR_30 group, while subjective and objective evaluation results in DPR_60 group and DPR_90 group increased (all P＜0.05) or being not significantly different from those in OSEM_120 group (all P＞0.05). No significant difference of liver SUVmean was found among 7 groups (P=0.955). SUVmax and SUVmean of lesions and livers obtained based on DPR_30,DPR_60 and DPR_90 groups were in good agreement with those oibtained based on OSEM_120 group. Conclusion Using DPR algorithm to reconstruct whole-body 18F-FDG PET image could shorten acquisition time under the premise of ensuring image quality. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
