基于模拟退火剂量计算模型的自动匹配算法

基于模拟退火优化算法，提出了一种点核卷积叠加剂量计算模型参数的自动匹配算法.模拟退火优化算法具有可以达到全局最优的优势，但模拟退火优化的随机性是不可控的.针对此问题，基于点核叠加剂量计算原理采用了定向约束模型能谱参数的方法，使得优化过程的迭代效率及稳定性大大提高.为了减少迭代次数，采用了BeamNRC软件模拟出加速器较好的初始能谱参数.通过临床实际加速器设备数据进行试验，结果表明，利用模拟退火优化算法进行模型参数的自动匹配的方法在保证临床精度要求的前提下，优化时间基本上控制在16h左右，在临床上是完全可以接受的，甚至针对一些加速器较好的能谱初始值，优化时间比国际上同类产品的优化时间还要短.模型参数的自动匹配功能也大大降低了软件对操作人员业务能力的依赖，增加了产品的安全性，降低了产品的维护成本.

A Method of Automatically Matching Dose Calculation Model Based on Simulated Annealing Arithmetic

Based on simulated annealing arithmetic， a method of automatically matching point kernel convolution/superposition dose calculation model was suggested. A global optimal solution can be reached with simulated annealing arithmetic， but it is uncontrollable because of its randomicity. To solve this problem， a method of limiting the energy spectrum sample space was used， which improved the optimization efficiency and stability. BeamNRC software was also used to generate a good initial energy spectrum in order to reduce iteration times. From results of the test on the machine data， it was concluded that the method mentioned in this article could meet the clinical requirements since the optimization time was around one point six hour. Having good initial energy spectrum， the result was better than the international similar products. The dependence degree of the operator technique and the maintenance cost could be reduced， and the product competitiveness could be improved with this method.