Pang Haowen, Qi Shouliang, Wu Yanan, Wang Meihuan, Li Chen, Sun Yu, Qian Wei, Tang Guoyan, Xu Jiaxuan, Liang Zhenyu, Chen Rongchang
College of Medicine and Biological Information Engineering, Northeastern University, Shenyang, China; Key Laboratory of Intelligent Computing in Medical Image, Ministry of Education, Northeastern University, Shenyang, China.
College of Medicine and Biological Information Engineering, Northeastern University, Shenyang, China; Key Laboratory of Intelligent Computing in Medical Image, Ministry of Education, Northeastern University, Shenyang, China.
Comput Methods Programs Biomed. 2023 Apr;231:107389. doi: 10.1016/j.cmpb.2023.107389. Epub 2023 Feb 2.
Non-contrast CT (NCCT) and contrast-enhanced CT (CECT) are important diagnostic tools with distinct features and applications for chest diseases. We developed two synthesizers for the mutual synthesis of NCCT and CECT and evaluated their applications.
Two synthesizers (S and S) were proposed based on a generative adversarial network. S generated synthetic CECT (SynCECT) from NCCT and S generated synthetic NCCT (SynNCCT) from CECT. A new training procedure for synthesizers was proposed. Initially, the synthesizers were pretrained using self-supervised learning (SSL) and dual-energy CT (DECT) and then fine-tuned using the registered NCCT and CECT images. Pulmonary vessel segmentation from NCCT was used as an example to demonstrate the effectiveness of the synthesizers. Two strategies (ST and ST) were proposed for pulmonary vessel segmentation. In ST, CECT images were used to train a segmentation model (Model-CECT), NCCT images were converted to SynCECT through S, and SynCECT was input to Model-CECT for testing. In ST, CECT data were converted to SynNCCT through S. SynNCCT and CECT-based annotations were used to train an additional model (Model-NCCT), and NCCT was input to Model-NCCT for testing. Three datasets, D1 (40 paired CTs), D2 (14 NCCTs and 14 CECTs), and D3 (49 paired DECTs), were used to evaluate the synthesizers and strategies.
For S, the mean absolute error (MAE), mean squared error (MSE), peak signal-to-noise ratio (PSNR), and structural similarity index (SSIM) were 14.60± 2.19, 1644± 890, 34.34± 1.91, and 0.94± 0.02, respectively. For S, they were 12.52± 2.59, 1460± 922, 35.08± 2.35, and 0.95± 0.02, respectively. Our synthesizers outperformed the counterparts of CycleGAN, Pix2Pix, and Pix2PixHD. The results of ablation studies on SSL pretraining, DECT pretraining, and fine-tuning showed that performance worsened (for example, for S, MAE increased to 16.53± 3.10, 17.98± 3.10, and 20.57± 3.75, respectively). Model-NCCT and Model-CECT achieved dice similarity coefficients (DSC) of 0.77 and 0.86 on D1 and 0.77 and 0.72 on D2, respectively.
The proposed synthesizers realized mutual and high-quality synthesis between NCCT and CECT images; the training procedures, including SSL pretraining, DECT pretraining, and fine-tuning, were critical to their effectiveness. The results demonstrated the usefulness of synthesizers for pulmonary vessel segmentation from NCCT images.
非增强CT（NCCT）和增强CT（CECT）是胸部疾病重要的诊断工具，具有不同的特点和应用。我们开发了两种用于NCCT和CECT相互合成的合成器，并评估了它们的应用。
基于生成对抗网络提出了两种合成器（S和S）。S从NCCT生成合成CECT（SynCECT），S从CECT生成合成NCCT（SynNCCT）。提出了一种合成器的新训练程序。最初，使用自监督学习（SSL）和双能CT（DECT）对合成器进行预训练，然后使用配准的NCCT和CECT图像进行微调。以NCCT的肺血管分割为例展示合成器的有效性。提出了两种肺血管分割策略（ST和ST）。在ST中，使用CECT图像训练分割模型（Model-CECT），通过S将NCCT图像转换为SynCECT，并将SynCECT输入Model-CECT进行测试。在ST中，通过S将CECT数据转换为SynNCCT。使用SynNCCT和基于CECT的标注训练另一个模型（Model-NCCT），并将NCCT输入Model-NCCT进行测试。使用三个数据集D1（40对CT）、D2（14个NCCT和14个CECT）和D3（49对DECT）评估合成器和策略。
对于S，平均绝对误差（MAE）、均方误差（MSE）、峰值信噪比（PSNR）和结构相似性指数（SSIM）分别为14.60±2.19、1644±890、34.34±1.91和0.94±0.02。对于S，它们分别为12.52±2.59、1460±922、35.08±2.35和0.95±0.02。我们的合成器优于CycleGAN、Pix2Pix和Pix2PixHD的同类产品。关于SSL预训练、DECT预训练和微调的消融研究结果表明性能变差（例如，对于S，MAE分别增加到16.53±3.10、17.98±3.10和20.57±3.75）。Model-NCCT和Model-CECT在D1上的骰子相似系数（DSC）分别为0.77和0.86，在D2上分别为0.77和0.72。
所提出的合成器实现了NCCT和CECT图像之间的相互高质量合成；包括SSL预训练、DECT预训练和微调在内的训练程序对其有效性至关重要。结果证明了合成器在从NCCT图像进行肺血管分割方面的有用性。
