目的 使用基于反向传播(BP)神经网络的机器学习方法对鼻咽癌患者的Gamma通过率建立预测模型,并评估其实用性.方法 收集2018-01-01-2020-10-31苏州大学附属第一医院放疗科收治的106例Ⅱ～Ⅲ期鼻咽癌患者临床资料,均采用同步加量9野均分动态调强治疗方案.提取患者放疗计划中与射野复杂性相关的31个指标,同时用map-check2测量954个射野的实际剂量分布,得到2％/2 mm、2％/3 mm、3％/2 mm和3％/3 mm(10％阈值和全局归一)4组参数下的Gamma通过率,将射野复杂度指标与4组Gamma通过率分别进行Spearman相关性分析,选择其中有统计学意义的指标作为神经网络学习的输入特征,90％数据作为训练集,10％数据作为测试集,对4组参数分别建立神经网络预测模型.结果 在与射野复杂性相关的31个指标中,调制复杂度、射野调制性等多个指标与测量Gamma通过率的Spearman相关系数差异有统计学意义,均P<0.05.2％/2 mm、2％/3 mm、3％/2 mm和3％/3 mm 4组神经网络模型预测的Gamma通过率最大偏差分别为4.10％、2.80％、3.20％和2.77％,其中99％的射野预测值与实际测量值的绝对偏差≤3％.结论 基于计划射野复杂性特征建立的BP神经网络模型可以对不同标准Gamma通过率进行准确预测,有助于提前发现Gamma通过率低的计划,对临床计划设计和放射治疗质量保证(QA)工作都具有一定的指导意义.

目的 利用锥形束CT(CBCT)研究宫颈癌动态容积弧形调强放疗的摆位误差大小,分析摆位误差对肿瘤计划靶体积(PTV)及危及器官受照剂量分布的影响.方法 应用医科达Synergy IGRT加速器对16例宫颈癌患者,行放疗前3次及以后每周1次的治疗前CBCT扫描,将获取的CBCT图像和计划CT图像进行匹配,获得左右(X)、头脚(Y)、腹背(Z)3个方向的摆位误差,将上述误差引入治疗计划系统重新计算得到PTV及危及器官的剂量分布参数并与原计划行配对t检验.结果 X、Y、Z方向上的摆位误差分别为(0.36±1.18)mm、(-0.11±3.45)mm、(1.19±1.37)mm;再计划的PTV 95％体积剂量(D95)明显小于原计划(P<0.05);再计划膀胱V45明显大于原计划(P<0.05);直肠V45与原计划相比无明显变化(P>0.05).结论 摆位误差导致PTV D95明显不足,危及器官直肠受照体积剂量增加,因此宫颈癌放疗过程中摆位误差的在线纠正是必要的.

目的 旨在研发出一套适用于放疗科日常管理的放射治疗申请单数据库.方法 基于常用的办公软件Windows Office,在Windows平台下,Infopath软件进行前端表单设计,Access软件进行后台数据库管理.结果 放射治疗申请单数据库能够实现医生对患者信息、计划信息的录入,物理师对数据库管理备份,权限分明,数据安全.结论 放射治疗申请单数据库操作简单,实用性强,运行稳定,是放疗科信息化建设的重要部分,能够推广应用.

Objective: To monitor the set-up error and the residual set-up error by weekly kilovolt cone-beam CT (KV-CBCT) in three-dimensional conformal radiotherapy for patients with esophageal cancer, and provide the basis for reasonable clinical target volume-planning target volume (CTV-PTV) margin.Methods: Nineteen patients with esophageal cancer receiving three-dimensional conformal radiotherapy were recruited in this study. The KV-CBCT scanning was conducted weekly for each patient, and the images in the corresponding three-dimensional match box were registrated with the positioning CT images. If the errors in three-dimensional directions [left-right (latitude, Lat), superior-inferior (vertical, Vrt) and anterior-posterior (longitude, Lng)] were more than threshold of 3 mm, an online reset of positioning should be performed. After that, the KV-CBCT scanning and then image registration were conducted again. The set-up errors after correction in three-dimensional directions were examined. The van Herk formula was used to calculate the theoretical CTV-PTV margin.Results: Before correction, the average set-up errors of 107 KV-CBCT images in Lat, Vrt and Lng directions were 0.39±0.31 cm, 0.24±0.23 cm, and 0.28±0.22 cm, respectively; altogether 79 (73.8%) set-up errors were more than threshold of 3 mm, resulting to an online correction. According to van Herk formula, the CTV-PTV margins before correction were 1.00, 0.91 and 0.96 cm, respectively. After correction, the average set-up errors of 107 KV-CBCT images in Lat, Vrt and Lng directions were 0.17±0.13 cm, 0.16±0.12 cm, and 0.14±0.14 cm, respectively. According to van Herk formula, the CTV-PTV margins after correction were 0.50, 0.46 and 0.49 cm,respectively.Conclusion: The set-up error can be reduced and the CTV-PTV margin can be narrowed through on-line KV-CBCT-guided radiation therapy. DOI:10.3781/j.issn.1000-7431.2015.77.331

Objective Objective To evaluate the set-up errors of intensity-modulated radiotherapy ( IMRT) in nasopha-ryngeal carcinoma ( NPC) and analyze its effects on dose distribution of targets and peripheral organs at risk ( OAR) using kilo-volt cone-beam CT ( KV-CBCT) .Methods 15 NPC patients who received IMRT were selected .They were set up by 3 points on thermoplastic film ,and scanned by KV-CBCT once a week .Sectional images were acquired after reconstruction and compared with the planning CT images ,the variations of three dimensional and horizontal rotation could be achieved .Transplanting the offset cen-tral point of three dimensional and horizontal rotation to the primary positioning images ,the other parameters were invariant .The dose distributions were simulated on the primary treatment planning CT images and the dose distributions betweem them were compared.Results The mean values of set-up errors on anteroposterior,craniocaudal,lateral position and rotation was -1.833 mm,-0.32 mm,0.96 mm,-0.8333 degree,respectively.The dose variations for GTVnxD95,GTVndD95,CTV2D95,PTV2D95,and PTVnd D95 were -1.28%~-0.32%,-3.31%~-0.63%,-2.33%~-1.30%,-2.51%~-0.45%,and -2.53%~-0.48%,respectively.Dmax of brainstem,spinal cord,and the optic chiasm were -3.14%~12.58%,-22.91%~21.21%,-7.25%~14.77%,respectively.The D50 and Dmean of parotid were -13.87%~18.21%and -6.84%~9.21%.Conclusion Due to the presences of set-up errors,dose distribution of nasopharyngeal carcinoma ( NPC) undergoing IMRT is affected ,and dose distribution of OAR increased significantly .KV-CBCT can reduce the treatment errors and improve the accuracy of IMRT .

Objective To analyze of the different accelerator dose rate conditions and the change in the treatment time and the number of MUs. Methods We used Varian's Eclipse Treatment Planning System to calculate the number of MUs and beam-on times for a total of 20 different treatment plans across an array of common IMRT sites including head and neck, chest, abdomen and central nervous system cancers using dose rates of 100, 200, 300, 400 and 600 MU/min. Results We observed a 4.5%~7%increase in the number of MUs delivered and decrease in the beam-on time for each 100 MU/min of dose rate increase. The exact magnitude of the changes depended on treatment site and planning target volume. Conclusion Although the number of MUs increased more rapidly for more complex treatment plans, the absolute beam-on time savings was greater for these plans because of the higher total number of MUs required to deliver them. We estimate that increasing the IMRT dose rate has the potential to treat more people per day for each linear accelerator.