子宫颈癌自适应放疗中关于个体化自动计划优化及放疗毒性预测的相关研究

Radiotherapy plays a crucial role in the cervical cancer treatment, but its target volume is always affected by bladder filling. Although saline injection or drinking instruction is used in clinical practice to maintain the bladder volume constant, urinary tract infection or poorly-filling bladder would be discovered, which would increase the risk of the under-dosing target or radiotherapy complications. To reduce the impact of bladder filling is challenging during cervical cancer treatment. In our previous controlled trial, we have proposed a bladder dynamic monitoring model and established a parameterized model for mean urinary inflow rate. These models could reduce the patient discomfort caused by holding onto the urine, as well as reduce technician labour and cost. In this work, we intend to establish the database of bladder filling to analyze the changes of three-dimension projections of bladder and then calculate the actual dose to target by this database; offline adaptive radiotherapy could be performed. In the bladder filling database combined with weekly cone beam CT (CBCT) images, the most similar radiotherapy plan will be chosen according to the target volume similarity index and target surface distribution. The parameters of the most similar plan will be used as the initial solution of the optimal plan and then real-time automatic optimization will be performed using mixed integer linear programming model to carry out online adaptive radiotherapy. The transmission detector will be used to execute the real-time in vivo dose monitoring during IMRT; the efficacy of radiotherapy and the incidence of secondary primary cancer will be evaluated, and the normal tissue complications of the new patients will be predicted, ultimately giving some prevention measures. Finally, the individual clinical pathway for the cervical cancer radiotherapy will be established preliminarily.

放疗在子宫颈癌治疗中起着关键性作用，但其靶区易受膀胱充盈的影响；尽管有注射盐水、指令涨尿等临床流程，但调查显示患者多出现尿路感染或膀胱充盈不佳，在高适形的调强放疗中更易造成靶区漏照或副反应加重。因此，如何减少膀胱影响是子宫颈癌治疗中亟待解决的问题。申请人目前建立了膀胱体积动态监测模式和总体/个体平均涨尿速率模型。本课题拟通过CT和锥形束CT（CBCT）图像结合患者的年龄、放疗计划等信息，建立膀胱充盈三维数据库，分析其变化规律，计算实际辐照剂量，开展离线自适应放疗；利用膀胱充盈三维数据库，结合每周CBCT图像，依据靶体积相似指数及表面分布等参数，寻找最相近的数据库计划，提取其参数作为优化初始解，使用混合整数线性规划模型，实时自动优化参数权重，开展在线自适应放疗；利用透射型探测器，开展患者在体剂量监测，评估正常组织放疗并发症和放疗致二次原位癌发生率；初步建立子宫颈癌体外放疗个体规范化临床路径。

如何减少膀胱充盈不一对靶区实际接受辐照剂量的影响及如何进行实时剂量传输监测是宫颈癌放疗中亟待解决的问题。尽管目前有些放疗前预处理措施，但临床实际中仍然无法完全消除大量残留的器官变化。本课题在前期提出的超声引导宫颈癌放疗个体化舒适性涨尿策略的基础上，开展了膀胱数据库建立、膀胱形态量化、剂量预测模型训练、实时剂量评估等系列研究，以期解决该问题。主要内容和结果如下：（1）通过CT和CBCT图像结合患者的年龄、放疗计划等信息，建立了膀胱充盈数据库，分析了膀胱体积变化规律，拟合了涨尿率Vtot计算公式，以此评估患者个体膀胱体积随时间变化；基于数据库，通过分析质心位置和三轴长度，量化了膀胱形态和运动变化，提出结合患者个体实际情况，可选择膀胱充盈的不同阶段（10−80ml、80−320ml、320−600mL、600−800ml）去评估膀胱空间变化；基于数据库，计算肿瘤和膀胱的实际辐照剂量，指导后续剂量传输；（2）利用CT影像及计划信息，基于U-Net网络训练进行宫颈癌剂量分布预测，预测值和计划值一致性较好，为建立基于预测剂量分布作为优化初始解的单一优化目标函数生成可执行计划提供了良好基础，为开展在线自适应提供有力保障；（3）自动计划能够通过辅助结构对靶区内的冷点及热点进行很好控制，均匀性指标比人工计划更优；但是，就适形指数而言，人工计划则更具优势，并且在膀胱和直肠的V40以及左、右侧股骨头的V30、V35等剂量参数上，人工计划也存在明显优势，这为如何开展个体在线自适应提供考虑；（4）验证了基于透射型探测器进行放疗中剂量实时传输监测的临床可行性，利用计算的修正因子，可准确评估射线传输至患者前的束流通量信息，结合组织轮廓，计算剂量-体积直方图（DVH）；PerFRACTIONTM结合γ通过率、剂量偏差（DD）等指标可监测治疗分次间加速器机械到位、射野跳数传输准确性及患者辐照剂量变化，可提高放疗安全性并为患者自适应放疗提供指导；（5）放疗并发症和放疗致二次原位癌的发生与器官被辐照剂量有关，对于接受首程放疗时相对年轻的长期存活者风险较高，预测结果需要进一步大数据验证。