Laboratory of Image Science and Technology, School of Computer Science and Engineering Southeast University, Nanjing, Jiangsu, P.R. China.
Department of Radiation Oncology Physics and Technology, Shandong Cancer Hospital and Institute, Shandong First Medical University and Shandong Academy of Medical Sciences, Jinan, P.R. China.
Med Phys. 2023 Dec;50(12):7779-7790. doi: 10.1002/mp.16566. Epub 2023 Jun 30.
The main application of [18F] FDG-PET ( FDG-PET) and CT images in oncology is tumor identification and quantification. Combining PET and CT images to mine pulmonary perfusion information for functional lung avoidance radiation therapy (FLART) is desirable but remains challenging.
To develop a deep-learning-based (DL) method to combine FDG-PET and CT images for producing pulmonary perfusion images (PPI).
Pulmonary technetium-99 m-labeled macroaggregated albumin SPECT (PPI ), FDG-PET, and CT images obtained from 53 patients were enrolled. CT and PPI images were rigidly registered, and registration displacement was subsequently used to align FDG-PET and PPI images. The left/right lung was separated and rigidly registered again to improve the registration accuracy. A DL model based on 3D Unet architecture was constructed to directly combine multi-modality FDG-PET and CT images for producing PPI (PPI ). 3D Unet architecture was used as the basic architecture, and the input was expanded from a single-channel to a dual-channel to combine multi-modality images. For comparative evaluation, FDG-PET images were also used alone to generate PPI . Sixty-seven samples were randomly selected for training and cross-validation, and 36 were used for testing. The Spearman correlation coefficient (r ) and multi-scale structural similarity index measure (MS-SSIM) between PPI /PPI and PPI were computed to assess the statistical and perceptual image similarities. The Dice similarity coefficient (DSC) was calculated to determine the similarity between high-/low- functional lung (HFL/LFL) volumes.
The voxel-wise r and MS-SSIM of PPI /PPI were 0.78 ± 0.04/0.57 ± 0.03, 0.93 ± 0.01/0.89 ± 0.01 for cross-validation and 0.78 ± 0.11/0.55 ± 0.18, 0.93 ± 0.03/0.90 ± 0.04 for testing. PPI /PPI achieved averaged DSC values of 0.78 ± 0.03/0.64 ± 0.02 for HFL and 0.83 ± 0.01/0.72 ± 0.03 for LFL in the training dataset and 0.77 ± 0.11/0.64 ± 0.12, 0.82 ± 0.05/0.72 ± 0.06 in the testing dataset. PPI yielded a stronger correlation and higher MS-SSIM with PPI than PPI (p < 0.001).
The DL-based method integrates lung metabolic and anatomy information for producing PPI and significantly improved the accuracy over methods based on metabolic information alone. The generated PPI can be applied for pulmonary perfusion volume segmentation, which is potentially beneficial for FLART treatment plan optimization.
[18F] FDG-PET（FDG-PET）和 CT 图像在肿瘤学中的主要应用是肿瘤识别和定量。结合 PET 和 CT 图像挖掘肺灌注信息以进行功能性肺回避放射治疗（FLART）是理想的，但仍然具有挑战性。
开发一种基于深度学习（DL）的方法，将 FDG-PET 和 CT 图像结合起来生成肺灌注图像（PPI）。
纳入了 53 名患者的肺锝-99 标记大聚合白蛋白 SPECT（PPI）、FDG-PET 和 CT 图像。对 CT 和 PPI 图像进行刚性配准，并随后使用配准位移来对齐 FDG-PET 和 PPI 图像。再次对左/右肺进行分离和刚性配准，以提高配准精度。构建了基于 3D Unet 架构的 DL 模型，直接将多模态 FDG-PET 和 CT 图像结合起来生成 PPI（PPI）。3D Unet 架构用作基本架构，输入从单通道扩展到双通道，以结合多模态图像。为了进行比较评估，还单独使用 FDG-PET 图像生成 PPI。随机选择 67 个样本进行训练和交叉验证，36 个样本用于测试。计算 PPI / PPI 与 PPI 之间的 Spearman 相关系数（r）和多尺度结构相似性指数度量（MS-SSIM），以评估统计和感知图像相似性。计算 Dice 相似系数（DSC）以确定高/低功能肺（HFL/LFL）体积之间的相似性。
PPI / PPI 的体素级 r 和 MS-SSIM 在交叉验证时分别为 0.78±0.04/0.57±0.03、0.93±0.01/0.89±0.01，在测试时分别为 0.78±0.11/0.55±0.18、0.93±0.03/0.90±0.04。PPI / PPI 在训练数据集中获得的平均 DSC 值为 0.78±0.03/0.64±0.02 用于 HFL 和 0.83±0.01/0.72±0.03 用于 LFL，在测试数据集中为 0.77±0.11/0.64±0.12、0.82±0.05/0.72±0.06。与 PPI 相比，基于 DL 的方法通过整合肺代谢和解剖信息来生成 PPI，与 PPI 相比具有更强的相关性和更高的 MS-SSIM（p<0.001）。
基于深度学习的方法整合了肺代谢和解剖信息来生成 PPI，与仅基于代谢信息的方法相比，显著提高了准确性。生成的 PPI 可用于肺灌注体积分割，这可能有利于 FLART 治疗计划优化。
