Cone-beam computed tomography (CBCT)- and magnetic resonance (MR)-images allow a daily observation of patient anatomy but are not directly suited for accurate proton dose calculations. This can be overcome by creating synthetic CTs (sCT) using deep convolutional neural networks. In this study, we compared sCTs based on CBCTs and MRs for head and neck (H&N) cancer patients in terms of image quality and proton dose calculation accuracy.

A dataset of 27 H&N-patients, treated with proton therapy (PT), containing planning CTs (pCTs), repeat CTs, CBCTs and MRs were used to train two neural networks to convert either CBCTs or MRs into sCTs. Image quality was quantified by calculating mean absolute error (MAE), mean error (ME) and Dice similarity coefficient (DSC) for bones. The dose evaluation consisted of a systematic non-clinical analysis and a clinical recalculation of actually used proton treatment plans. Gamma analysis was performed for non-clinical and clinical treatment plans. For clinical treatment plans also dose to targets and organs at risk (OARs) and normal tissue complication probabilities (NTCP) were compared.

CBCT-based sCTs resulted in higher image quality with an average MAE of 40 4 HU and a DSC of 0.95, while for MR-based sCTs a MAE of 65 4 HU and a DSC of 0.89 was observed. Also in clinical proton dose calculations, sCT_CBCT achieved higher average gamma pass ratios (2%/2 mm criteria) than sCT_MR (96.1% vs. 93.3%). Dose-volume histograms for selected OARs and NTCP-values showed a very small difference between sCT_CBCT and sCT_MR and a high agreement with the reference pCT.

CBCT- and MR-based sCTs have the potential to enable accurate proton dose calculations valuable for daily adaptive PT. Significant image quality differences were observed but did not affect proton dose calculation accuracy in a similar manner. Especially the recalculation of clinical treatment plans showed high agreement with the pCT for both sCT_CBCT and sCT_MR.

锥形束计算机断层扫描 (CBCT) 和磁共振 (MR) 图像允许每天观察患者解剖结构，但不直接适用于准确的质子剂量计算。这可以通过使用深度卷积神经网络创建合成 CT (sCT) 来克服。在本研究中，我们在图像质量和质子剂量计算准确性方面比较了基于 CBCT 和 MR 的头颈部 (H&N) 癌症患者的 sCT。

使用质子治疗 (PT) 治疗的 27 名 H&N 患者的数据集，其中包含计划 CT (pCT)、重复 CT、CBCT 和 MR，用于训练两个神经网络以将 CBCT 或 MR 转换为 sCT。通过计算骨骼的平均绝对误差 (MAE)、平均误差 (ME) 和骰子相似系数 (DSC) 来量化图像质量。剂量评估包括系统的非临床分析和实际使用的质子治疗计划的临床重新计算。对非临床和临床治疗计划进行伽玛分析。对于临床治疗计划，还比较了目标和风险器官 (OAR) 和正常组织并发症概率 (NTCP) 的剂量。

基于 CBCT 的 sCT 的图像质量更高，平均 MAE 为 40 4 HU，DSC 为 0.95，而基于 MR 的 sCT 的 MAE 为 65 4 HU，DSC 为 0.89。同样在临床质子剂量计算中，sCT _CBCT的平均伽马通过率（2%/2 mm 标准）高于 sCT _MR（96.1% 对 93.3%）。_{所选 OAR 和 NTCP 值的剂量体积直方图显示 sCT CBCT}和 sCT _MR之间的差异非常小，并且与参考 pCT 高度一致。

基于 CBCT 和 MR 的 sCT 有可能实现对日常自适应 PT 有价值的准确质子剂量计算。观察到显着的图像质量差异，但不会以类似的方式影响质子剂量计算的准确性。_{尤其是临床治疗计划的重新计算与sCT CBCT}和sCT _MR的pCT高度一致。