论文部分内容阅读
目的:比较遗传算法优化的误差反向传播算法(GA-BP)和支持向量机(SVM)对于呼吸运动估计的效果,为临床放疗中的呼吸运动估计提供参考。方法:在MATLAB平台下分别采用GA-BP算法和SVM对德国吕贝克大学机器人与认知系统研究所公开的两组可免费下载的实验数据进行计算,并对两种算法的估计精度和实时性进行比较分析。结果:在GA-BP算法中,测试数据中90%的估计结果均方根误差小于1.09 mm;在SVM中,90%的估计结果均方根误差小于1.72 mm。两种算法单次估计的时间都小于0.3 ms,均满足临床使用需求。结论:在呼吸运动估计中,两种算法均能满足临床实时性需求,且GA-BP算法估计结果的均方根误差小于SVM。
OBJECTIVE: To compare the effects of GA-BP and SVM on respiratory motion estimation by genetic algorithm, and provide a reference for respiratory motion estimation in clinical radiotherapy. Methods: The GA-BP algorithm and SVM were used respectively to calculate the experimental data freely available for download published by Robotics and Cognitive Systems Institute of Lübeck University in Germany. The accuracy and real-time performance of the two algorithms were evaluated. For comparative analysis. Results: In the GA-BP algorithm, the root mean square error of 90% of the estimated results in the test data is less than 1.09 mm. In the SVM, the root mean square error of 90% of the estimated results is less than 1.72 mm. The single estimation time of the two algorithms is less than 0.3 ms, both meet the clinical needs. Conclusion: In the respiratory motion estimation, the two algorithms can meet the real-time clinical needs, and the root mean square error of the GA-BP algorithm estimation results is less than the SVM.