Department of Radiation Oncology, Peking University Third Hospital, Beijing, China.
School of Biomedical Engineering, Capital Medical University, Beijing, China.
Med Phys. 2022 Mar;49(3):1522-1534. doi: 10.1002/mp.15460. Epub 2022 Jan 27.
Cone-beam computed tomography (CBCT) is frequently used for accurate image-guided radiation therapy. However, the poor CBCT image quality prevents its further clinical use. Thus, it is important to improve the HU accuracy and structure preservation of CBCT images.
In this study, we proposed a novel method to generate synthetic CT (sCT) images from CBCT images. A multiresolution residual deep neural network (RDNN) was adopted for image regression from CBCT images to planning CT (pCT) images. At the coarse level, RDNN was first trained with a large amount of lower resolution images, which can make the network focus on coarse information and prevent overfitting problems. More fine information was obtained gradually by fine-tuning the coarse model using fewer number of higher resolution images. Our model was optimized by using aligned pCT and CBCT image pairs of a particular body region of 153 prostate cancer patients treated in our hospital (120 for training and 33 for testing). Five-fold cross-validation was used to tune the hyperparameters and the testing data were used to evaluate the performance of the final models.
The mean absolute error (MAE) between CBCT and pCT on the testing data was 352.56 HU, while the MAE between the sCT and pCT images was 52.18 HU for our proposed multiresolution RDNN model, which reduced the MAE by 85.20% (p < 0.01). In addition, the average structural similarity index measure between the sCT and CBCT was 19.64% (p = 0.01) higher than that of pCT and CBCT.
The sCT images generated using our proposed multiresolution RDNN have higher HU accuracy and structural fidelity, which may promote the further applications of CBCT images in the clinic for structure segmentation, dose calculation, and adaptive radiotherapy planning.
锥形束计算机断层扫描（CBCT）常用于精确的图像引导放射治疗。然而，较差的 CBCT 图像质量阻止了其进一步的临床应用。因此，提高 CBCT 图像的 HU 准确性和结构保真度非常重要。
在这项研究中，我们提出了一种从 CBCT 图像生成合成 CT（sCT）图像的新方法。采用多分辨率残差深度神经网络（RDNN）从 CBCT 图像到计划 CT（pCT）图像进行图像回归。在粗尺度上，首先使用大量低分辨率图像对 RDNN 进行训练，这可以使网络专注于粗信息，并防止过拟合问题。通过使用较少数量的高分辨率图像对粗模型进行微调，逐渐获得更多的精细信息。我们的模型使用 153 名在我院接受治疗的前列腺癌患者特定身体区域的 pCT 和 CBCT 图像对（120 个用于训练，33 个用于测试）进行优化。使用五折交叉验证来调整超参数，使用测试数据来评估最终模型的性能。
在测试数据中，CBCT 与 pCT 之间的平均绝对误差（MAE）为 352.56 HU，而我们提出的多分辨率 RDNN 模型生成的 sCT 与 pCT 图像之间的 MAE 为 52.18 HU，降低了 85.20%（p < 0.01）。此外，sCT 与 CBCT 之间的平均结构相似性指数测量值比 pCT 与 CBCT 高 19.64%（p = 0.01）。
使用我们提出的多分辨率 RDNN 生成的 sCT 图像具有更高的 HU 准确性和结构保真度，这可能会促进 CBCT 图像在临床结构分割、剂量计算和自适应放射治疗计划中的进一步应用。
