肿瘤放疗并发症概率预测模型参数拟合方法

发布时间：2018-02-15 01:37

本文关键词： 肿瘤放射治疗并发症 NTCP模型　出处：《东南大学学报(自然科学版)》2015年02期 　论文类型：期刊论文

【摘要】：为了建立具有群体特异性的肿瘤放疗NTCP预测模型,提出了一种模型参数拟合方法.首先,基于NTCP模型的特点构建最大似然函数;然后,分别采用确定性优化方法和随机性优化方法对最大似然函数进行优化,分析优化过程的时间成本及优化结果,探讨用于拟合NTCP模型参数的最优方法.实验结果表明,用于拟合NTCP模型参数的最大似然函数是非凸的,存在局部最优解;遗传算法是一种最稳定的最大似然函数优化方法,其运行时间比模拟退火算法短,而且可以在每次优化结束后给出全局最优解,以作为NTCP模型参数.所提方法可以帮助肿瘤放疗工作者在临床随访数据的基础上建立具有群体特异性的放疗并发症预测模型.
[Abstract]:In order to establish a population-specific NTCP prediction model for tumor radiotherapy, a model parameter fitting method is proposed. Firstly, the maximum likelihood function is constructed based on the characteristics of NTCP model, and then the maximum likelihood function is constructed. The deterministic optimization method and stochastic optimization method are used to optimize the maximum likelihood function respectively. The time cost and optimization results of the optimization process are analyzed, and the optimal method for fitting the parameters of the NTCP model is discussed. The experimental results show that, The maximum likelihood function used to fit the parameters of NTCP model is nonconvex and has a local optimal solution. Genetic algorithm is the most stable maximum likelihood function optimization method, and its running time is shorter than that of simulated annealing algorithm. After each optimization, the global optimal solution can be given as a parameter of NTCP model. The proposed method can help tumor radiotherapy workers to establish a group specific prediction model of radiotherapy complications on the basis of clinical follow-up data.
【作者单位】：东南大学影像科学与技术实验室;山东省肿瘤防治研究院;雷恩第一大学信号与图像实验室;
【基金】：国家自然科学基金资助项目(61271312,81272501,81301298) 国家重点基础研究发展计划(973计划)资助项目(2011CB707904)
【分类号】：R730.55

【参考文献】