摘要
获取高强度聚焦超声(High Intensity Focused Ultrasound,HIFU)辐照前后的新鲜离体猪肉组织减影图像,通过二维经验模态分解(Bidimensional Empirical Mode Decomposition,BEMD)方法将其分解为两个固有模态函数(IMF)和一个余量.将各分量图像的灰度均值作为反应温度变化的解释变量构建HIFU治疗无损测温的随机森林(RF)模型,对图像温度进行预测.对比分析同样以灰度均值作为解释变量的预测B超图像温度的线性回归方程模型、支持向量机(SVM)模型、BEMD-SVM模型以及RF模型的预测结果,发现BEMD-RF模型测温有着较高的精度,在40～75℃的温度范围内,测温误差在3℃以内,满足HIFU治疗过程中对温度监测的需求,实用性更强.
Irradiating fresh pork in vitro by high intensity focused ultrasound,subtraction image of B ultrasonic images were obtained before and after irradiation.The subtraction image was decomposed into two Intrinsic Mode Functions and a residue by Bidimensional Empirical Mode Decomposition(BEMD).The random forest(RF)model for HIFU therapeutic non-invasive temperature measurement was constructedby using the gray level mean of each component image as an explanatory variable of reflecting temperature change,and the image temperature was predicted.The linear regression equation model,the support vector machine(SVM)model,the BEMD-SVM model and the RF model are also compared for predicting B-mode image temperature with the mean value of gray scale as the explanatory variable.The results show that the BEMD-RF model has a high temperature measurement accuracy.Within the temperature range of 40℃~75℃,the temperature measurement error is within 3℃,which meets the needs of temperature monitoring during HIFU treatment.
作者
郭燕
丁亚军
钱盛友
陈兴
GUO Yan;DING Yajun;QIAN Shengyou;CHEN Xing(College of Information Science and Engineering,Hunan Normal University,Changsha 410081,China;School of Physics and Electronics,Hunan Normal University,Changsha 410081,China)
出处
《测试技术学报》
2018年第6期487-492,共6页
Journal of Test and Measurement Technology
基金
国家自然科学基金资助项目(11474090
11774088)
湖南省自然科学基金资助项目(2016JJ3090)
关键词
高强度聚焦超声
二维经验模态分解
无损测温
随机森林
支持向量机
high intensity focused ultrasound
Bidimensional Empirical Mode Decomposition
non-invasive temperature measurement
random forest
support vector machine
