Quantification of ¹⁸F-FDG PET images is useful for accurate diagnosis and evaluation of various brain diseases, including brain tumors, epilepsy, dementia, and Parkinson disease. However, accurate quantification of ¹⁸F-FDG PET images requires matched 3-dimensional T₁ MRI scans of the same individuals to provide detailed information on brain anatomy. In this paper, we propose a transfer learning approach to adapt a pretrained deep neural network model from amyloid PET to spatially normalize ¹⁸F-FDG PET images without the need for 3-dimensional MRI. Methods: The proposed method is based on a deep learning model for automatic spatial normalization of ¹⁸F-FDG brain PET images, which was developed by fine-tuning a pretrained model for amyloid PET using only 103 ¹⁸F-FDG PET and MR images. After training, the algorithm was tested on 65 internal and 78 external test sets. All T₁ MR images with a 1-mm isotropic voxel size were processed with FreeSurfer software to provide cortical segmentation maps used to extract a ground-truth regional SUV ratio using cerebellar gray matter as a reference region. These values were compared with those from spatial normalization-based quantification methods using the proposed method and statistical parametric mapping software. Results: The proposed method showed superior spatial normalization compared with statistical parametric mapping, as evidenced by increased normalized mutual information and better size and shape matching in PET images. Quantitative evaluation revealed a consistently higher SUV ratio correlation and intraclass correlation coefficients for the proposed method across various brain regions in both internal and external datasets. The remarkably good correlation and intraclass correlation coefficient values of the proposed method for the external dataset are noteworthy, considering the dataset’s different ethnic distribution and the use of different PET scanners and image reconstruction algorithms. Conclusion: This study successfully applied transfer learning to a deep neural network for ¹⁸F-FDG PET spatial normalization, demonstrating its resource efficiency and improved performance. This highlights the efficacy of transfer learning, which requires a smaller number of datasets than does the original network training, thus increasing the potential for broader use of deep learning–based brain PET spatial normalization techniques for various clinical and research radiotracers.

¹⁸ F-FDG PET 图像的量化有助于准确诊断和评估各种脑部疾病，包括脑肿瘤、癫痫、痴呆和帕金森病。然而， ¹⁸ F-FDG PET 图像的准确量化需要对同一个体进行匹配的 3 维 T ₁ MRI 扫描，以提供有关大脑解剖结构的详细信息。在本文中，我们提出了一种迁移学习方法，以适应来自淀粉样蛋白 PET 的预训练深度神经网络模型，以在不需要 3 维 MRI 的情况下对¹⁸ F-FDG PET 图像进行空间标准化。方法：所提出的方法基于深度学习模型，用于¹⁸ F-FDG 脑部 PET 图像的自动空间归一化，该模型是通过仅使用 103 ¹⁸ F-FDG PET 和 MR 图像对淀粉样蛋白 PET 的预训练模型进行微调而开发的。训练结束后，该算法在 65 个内部测试集和 78 个外部测试集上进行了测试。所有具有 1 毫米各向同性体素尺寸的 T ₁ MR 图像均使用 FreeSurfer 软件进行处理，以提供皮质分割图，用于使用小脑灰质作为参考区域来提取地面真实区域 SUV 比率。使用所提出的方法和统计参数映射软件将这些值与基于空间归一化的量化方法的值进行比较。结果：与统计参数映射相比，所提出的方法显示出优越的空间归一化效果，这可以通过 PET 图像中归一化互信息的增加以及更好的尺寸和形状匹配来证明。 定量评估显示，所提出的方法在内部和外部数据集中的各个大脑区域具有一致较高的 SUV 比率相关性和类内相关系数。考虑到数据集的不同种族分布以及不同 PET 扫描仪和图像重建算法的使用，所提出的方法对于外部数据集的非常好的相关性和类内相关系数值值得注意。结论：本研究成功地将迁移学习应用于¹⁸ F-FDG PET 空间归一化的深度神经网络，展示了其资源效率和改进的性能。这凸显了迁移学习的功效，与原始网络训练相比，迁移学习所需的数据集数量更少，从而增加了基于深度学习的大脑 PET 空间标准化技术在各种临床和研究放射性示踪剂中更广泛使用的潜力。