剂量预测引导的鼻咽癌放疗计划质量定量评估方法

Dose prediction-guided quantitative assessment of the quality radiotherapy plan for nasopharyngeal cancer

目的:利用人工神经网络模型对鼻咽癌计划进行剂量预测,根据剂量预测值建立计划质量测度标准(PQM),实现放疗计划个性化定量评估。方法:回顾性分析于福建省肿瘤医院进行放射治疗的鼻咽癌计划114例。提取25项危及器官(OAR)与靶区之间的几何空间关系特征,训练(81例)得到基于人工神经网络的剂量预测模型并测试(23例)验证其准确性。分别用基于剂量限值建立PQM和基于剂量预测值建立PQM两种评估方法对10例临床通过计划进行定量评估,讨论两种评估方法的合理性。结果:11项鼻咽癌主要OAR剂量学指标,预测值与实际值的剂量相关总体平均差值为(-0.07±4.55)Gy,体积相关总体平均差值为-1.06%±3.80%,预测准确性可达90%。10例临床通过的鼻咽癌计划,基于剂量限值建立PQM方法对病例4评估为不合格,其余病例合格。基于剂量预测值建立PQM方法对病例9评估为不合格,其余病例合格。结论:剂量预测引导的鼻咽癌计划质量定量评估方法可以反映计划是否存在继续优化的空间,且该评估方法克服了剂量限值未考虑病例特异性的缺陷,能更科学合理地对放疗计划进行定量评估。

Objective To carry out the dose prediction for nasopharyngeal carcinoma radiotherapy using artificial neural network model,and to establish plan quality metrics(PQM)based on dose prediction for achieving personalized quantitative assessment of radiotherapy plan quality.Methods A total of 114 cases of nasopharyngeal carcinoma treated with radiotherapy in Fujian Cancer Hospital were analyzed retrospectively.After extracting 25 geometric spatial relationship features between organs-at-risk and target area,81 cases were used for training and obtaining an artificial neural network-based dose prediction model,and the other 23 cases were used for validating the accuracy of the prediction model.PQM were developed based on dose limits and dose prediction,separately,for the quantitative assessment of 10 clinical cases,and finally,the rationalities of the two methods were discussed.Results For the 11 main OAR dosimetric indicators of nasopharyngeal carcinoma,the overall mean difference between predictive and actual values was(-0.07±4.55)Gy for dose-related indicators and(-1.06±3.80)%for volume-related indicators;and the prediction accuracy reached 90%.For the 10 clinically approved nasopharyngeal carcinoma plans,the method of establishing PQM based on dose limits was assessed as failing for case 4 and passing for the remaining cases,while the method of establishing PQM based on dose prediction was assessed as failing for case 9 and passing for the remaining cases.Conclusion The dose prediction-guided quantitative assessment of the quality radiotherapy plan for nasopharyngeal cancer can reflect whether the plan can be further optimized.Moreover,the dose prediction-guided method overcomes the shortcomings of the dose limitsbased assessment which do not take case specificity into account,and can evaluate the radiotherapy plan more scientifically and reasonably.