Department of Radiology and Biomedical Imaging, Yale University, New Haven, CT, United States of America. Department of Engineering Physics, Tsinghua University, Beijing, People's Republic of China. Key Laboratory of Particle and Radiation Imaging, Ministry of Education (Tsinghua University), Beijing, People's Republic of China.
Phys Med Biol. 2019 Aug 21;64(16):165019. doi: 10.1088/1361-6560/ab3242.
Reducing radiation dose is important for PET imaging. However, reducing injection doses causes increased image noise and low signal-to-noise ratio (SNR), subsequently affecting diagnostic and quantitative accuracies. Deep learning methods have shown a great potential to reduce the noise and improve the SNR in low dose PET data. In this work, we comprehensively investigated the quantitative accuracy of small lung nodules, in addition to visual image quality, using deep learning based denoising methods for oncological PET imaging. We applied and optimized an advanced deep learning method based on the U-net architecture to predict the standard dose PET image from 10% low-dose PET data. We also investigated the effect of different network architectures, image dimensions, labels and inputs for deep learning methods with respect to both noise reduction performance and quantitative accuracy. Normalized mean square error (NMSE), SNR, and standard uptake value (SUV) bias of different nodule regions of interest (ROIs) were used for evaluation. Our results showed that U-net and GAN are superior to CAE with smaller SUV and SUV bias at the expense of inferior SNR. A fully 3D U-net has optimal quantitative performance compared to 2D and 2.5D U-net with less than 15% SUV bias for all the ten patients. U-net outperforms Residual U-net (r-U-net) in general with smaller NMSE, higher SNR and lower SUV bias. Fully 3D U-net is superior to several existing denoising methods, including Gaussian filter, anatomical-guided non-local mean (NLM) filter, and MAP reconstruction with Quadratic prior and relative difference prior, in terms of superior image quality and trade-off between noise and bias. Furthermore, incorporating aligned CT images has the potential to further improve the quantitative accuracy in multi-channel U-net. We found the optimal architectures and parameters of deep learning based methods are different for absolute quantitative accuracy and visual image quality. Our quantitative results demonstrated that fully 3D U-net can both effectively reduce image noise and control bias even for sub-centimeter small lung nodules when generating standard dose PET using 10% low count down-sampled data.
降低辐射剂量对于 PET 成像很重要。然而，降低注射剂量会导致图像噪声增加和信噪比（SNR）降低，从而影响诊断和定量准确性。深度学习方法在降低低剂量 PET 数据的噪声和提高 SNR 方面显示出巨大的潜力。在这项工作中，我们综合研究了基于深度学习的去噪方法对肿瘤 PET 成像中小肺结节的定量准确性，除了视觉图像质量。我们应用并优化了一种基于 U-net 架构的先进深度学习方法，从 10%低剂量 PET 数据中预测标准剂量 PET 图像。我们还研究了不同的网络架构、图像尺寸、标签和输入对深度学习方法在降噪性能和定量准确性方面的影响。不同感兴趣区（ROI）的归一化均方误差（NMSE）、SNR 和标准摄取值（SUV）偏差用于评估。我们的结果表明，U-net 和 GAN 优于 CAE，具有更小的 SUV 和 SUV 偏差，而 SNR 较低。与 2D 和 2.5D U-net 相比，全 3D U-net 具有最佳的定量性能，所有 10 名患者的 SUV 偏差均小于 15%。与 r-U-net 相比，U-net 在一般情况下具有更小的 NMSE、更高的 SNR 和更低的 SUV 偏差。全 3D U-net 在图像质量和噪声与偏差之间的权衡方面优于几种现有的去噪方法，包括高斯滤波器、解剖引导非局部均值（NLM）滤波器以及具有二次先验和相对差先验的 MAP 重建。此外，结合配准 CT 图像有可能进一步提高多通道 U-net 的定量准确性。我们发现，对于绝对定量准确性和视觉图像质量，基于深度学习的方法的最佳架构和参数是不同的。我们的定量结果表明，即使使用 10%低计数下采样数据生成标准剂量 PET，全 3D U-net 也可以有效地降低图像噪声并控制偏差，即使是对于亚厘米级的小肺结节。
