基于PCA的脂肪肝超声RF信号特征选择
|
摘要
随着对脂肪肝超声图像识别的深入研究,越来越多的识别特征被提出来。而特征之间的相关性造成干扰信息,使得识别率反而下降。文中基于超声射频信号提取特征参数,通过两独立样本均数的t检验和主成分分析法,从10个特征参数中,组合前三个特征值的特征向量得到组合特征。前三个特征值的累积贡献率达97.86,在去冗余的同时,保留了绝大部分的原始信息。将改进方法应用于脂肪肝超声图像的识别,平均识别率从选用所有特征的75.63提高到了88.99。
With the deep research on fatty liver ultrasound image recognition,more and more recognition characteristics were proposed.The interference information caused by the correlation of features makes the identification rate dropped. Therefore,how to select the appropriate features or the combination of features became a difficult point in research. This paper extracted characteristic parameters based on ultrasonic RF signal and got a combination feature by the combination of the feature vectors of the first three eigenvalues from10 characteristics,using t- test and principal component analysis. The cumulative contribution rate of the first three eigenvalues was97. 86,which removed redundancy while retain most of the original information. While the improved method was applied to the identification of fatty liver ultrasound images,the average recognition rate increased to 88. 99 from 75. 63 of all features selected.
With the deep research on fatty liver ultrasound image recognition,more and more recognition characteristics were proposed.The interference information caused by the correlation of features makes the identification rate dropped. Therefore,how to select the appropriate features or the combination of features became a difficult point in research. This paper extracted characteristic parameters based on ultrasonic RF signal and got a combination feature by the combination of the feature vectors of the first three eigenvalues from10 characteristics,using t- test and principal component analysis. The cumulative contribution rate of the first three eigenvalues was97. 86,which removed redundancy while retain most of the original information. While the improved method was applied to the identification of fatty liver ultrasound images,the average recognition rate increased to 88. 99 from 75. 63 of all features selected.
引文
[1]林江莉,汪小毅,李德玉,等.脂肪肝B超图像特征提取研究[J].四川大学学报:工程科学版,2005,37(1):130-134.
[2]Li Guokao,Luo Yu,Deng Wei,et al.Computer aided diagnosis of fatty liver ultrasonic images based on support vector machine[C].30th Annual International IEEE EMBS Conference.[s.l.]:IEEE Press,2008:4768-4771.
[3]Xie Xiuqun,Luo Yan,Quan Jierong,et al.SD rats’fatty liver tissue classification based on radiofrequency signal[C].Jin D,Lin S.Advances in Computer Science and Information Engineering.[s.l.]:Springer Press,2012:643-647.
[4]刘志东,罗燕,林江莉,等.基于超声射频RF信号的脂肪肝分级量化方法[J].四川大学学报:工程科学版,2011,43(Z1):160-164.
[5]Liu Lanbo,John W Lane b,Quan Youli.Radar attenuation tomography using the centroid frequency downshift method[J].Journal of Applied Geophysics,1988,40(1-3):105-116.
[6]徐克学.生物数学[M].北京:科学出版社,1999:51-70.
[7]刘茂福,胡慧君,何炎祥.主成分分析在图像Zernike矩特征降维中的应用[J].计算机应用,2007,27(3):696-698,702.
[8]刘小凤,李文.基于主成分分析的人脸特征提取[J].企业家天地:理论版,2010(12):245-246.
[9]亓文永,叶心太.应用于图像特征识别的主成分分析算法[J].科技信息,2008(22):30,19.
[10]周达丽.基于主成分分析的特征脸提取及人脸识别实现[J].软件导刊,2012,11(5):162-164.
[2]Li Guokao,Luo Yu,Deng Wei,et al.Computer aided diagnosis of fatty liver ultrasonic images based on support vector machine[C].30th Annual International IEEE EMBS Conference.[s.l.]:IEEE Press,2008:4768-4771.
[3]Xie Xiuqun,Luo Yan,Quan Jierong,et al.SD rats’fatty liver tissue classification based on radiofrequency signal[C].Jin D,Lin S.Advances in Computer Science and Information Engineering.[s.l.]:Springer Press,2012:643-647.
[4]刘志东,罗燕,林江莉,等.基于超声射频RF信号的脂肪肝分级量化方法[J].四川大学学报:工程科学版,2011,43(Z1):160-164.
[5]Liu Lanbo,John W Lane b,Quan Youli.Radar attenuation tomography using the centroid frequency downshift method[J].Journal of Applied Geophysics,1988,40(1-3):105-116.
[6]徐克学.生物数学[M].北京:科学出版社,1999:51-70.
[7]刘茂福,胡慧君,何炎祥.主成分分析在图像Zernike矩特征降维中的应用[J].计算机应用,2007,27(3):696-698,702.
[8]刘小凤,李文.基于主成分分析的人脸特征提取[J].企业家天地:理论版,2010(12):245-246.
[9]亓文永,叶心太.应用于图像特征识别的主成分分析算法[J].科技信息,2008(22):30,19.
[10]周达丽.基于主成分分析的特征脸提取及人脸识别实现[J].软件导刊,2012,11(5):162-164.