Division of Nuclear Medicine and Molecular Imaging, Geneva University Hospital, Geneva, Switzerland.
Persian Gulf Nuclear Medicine Research Center, Department of Molecular Imaging and Radionuclide Therapy (MIRT), Bushehr Medical University Hospital, Faculty of Medicine, Bushehr University of Medical Sciences, Bushehr, Iran.
Neuroimage. 2021 Dec 15;245:118697. doi: 10.1016/j.neuroimage.2021.118697. Epub 2021 Nov 4.
Reducing the injected activity and/or the scanning time is a desirable goal to minimize radiation exposure and maximize patients' comfort. To achieve this goal, we developed a deep neural network (DNN) model for synthesizing full-dose (FD) time-of-flight (TOF) bin sinograms from their corresponding fast/low-dose (LD) TOF bin sinograms.
Clinical brain PET/CT raw data of 140 normal and abnormal patients were employed to create LD and FD TOF bin sinograms. The LD TOF sinograms were created through 5% undersampling of FD list-mode PET data. The TOF sinograms were split into seven time bins (0, ±1, ±2, ±3). Residual network (ResNet) algorithms were trained separately to generate FD bins from LD bins. An extra ResNet model was trained to synthesize FD images from LD images to compare the performance of DNN in sinogram space (SS) vs implementation in image space (IS). Comprehensive quantitative and statistical analysis was performed to assess the performance of the proposed model using established quantitative metrics, including the peak signal-to-noise ratio (PSNR), structural similarity index metric (SSIM) region-wise standardized uptake value (SUV) bias and statistical analysis for 83 brain regions.
SSIM and PSNR values of 0.97 ± 0.01, 0.98 ± 0.01 and 33.70 ± 0.32, 39.36 ± 0.21 were obtained for IS and SS, respectively, compared to 0.86 ± 0.02and 31.12 ± 0.22 for reference LD images. The absolute average SUV bias was 0.96 ± 0.95% and 1.40 ± 0.72% for SS and IS implementations, respectively. The joint histogram analysis revealed the lowest mean square error (MSE) and highest correlation (R = 0.99, MSE = 0.019) was achieved by SS compared to IS (R = 0.97, MSE= 0.028). The Bland & Altman analysis showed that the lowest SUV bias (-0.4%) and minimum variance (95% CI: -2.6%, +1.9%) were achieved by SS images. The voxel-wise t-test analysis revealed the presence of voxels with statistically significantly lower values in LD, IS, and SS images compared to FD images respectively.
The results demonstrated that images reconstructed from the predicted TOF FD sinograms using the SS approach led to higher image quality and lower bias compared to images predicted from LD images.
降低注射剂量和/或扫描时间是降低辐射暴露和提高患者舒适度的理想目标。为了实现这一目标，我们开发了一种深度神经网络（DNN）模型，用于从相应的快速/低剂量（LD）TOF -bin 正弦图合成全剂量（FD）TOF-bin 正弦图。
使用 140 名正常和异常患者的临床脑 PET/CT 原始数据来创建 LD 和 FD TOF-bin 正弦图。LD TOF 正弦图是通过 FD 列表模式 PET 数据的 5%欠采样创建的。TOF 正弦图被分为七个时间 bin（0、±1、±2、±3）。分别使用残差网络（ResNet）算法来生成 FD bins 从 LD bins。还训练了一个额外的 ResNet 模型，用于从 LD 图像合成 FD 图像，以比较 DNN 在正弦图空间（SS）和图像空间（IS）中的性能。使用已建立的定量指标，包括峰值信噪比（PSNR）、结构相似性指数度量（SSIM）区域标准化摄取值（SUV）偏差和 83 个脑区的统计分析，对所提出的模型进行了全面的定量和统计分析。
IS 和 SS 的 SSIM 和 PSNR 值分别为 0.97±0.01、0.98±0.01 和 33.70±0.32、39.36±0.21，而参考 LD 图像的 SSIM 和 PSNR 值分别为 0.86±0.02 和 31.12±0.22。SS 和 IS 实现的绝对平均 SUV 偏差分别为 0.96±0.95%和 1.40±0.72%。联合直方图分析显示，与 IS（R=0.97，MSE=0.028）相比，SS 实现的均方误差（MSE）最低且相关性最高（R=0.99，MSE=0.019）。Bland & Altman 分析显示，SS 图像的 SUV 偏差最低（-0.4%），方差最小（95%CI：-2.6%，+1.9%）。体素-wise t 检验分析显示，与 FD 图像相比，LD、IS 和 SS 图像中存在统计学上显著较低值的体素。
结果表明，使用 SS 方法从预测的 TOF FD 正弦图重建图像导致更高的图像质量和更低的偏差，与从 LD 图像预测的图像相比。
