基于超声弹性成像的支持向量机对颈动脉易损斑块的自动识别
|
摘要
本研究的目的是使用基于超声弹性成像的支持向量机(Support Vector Machine,SVM)实现对颈动脉易损斑块的自动检测。共采集了52个志愿者的80例颈动脉粥样硬化斑块的超声长轴射频数据,利用弹性成像算法得到应变率分布,并提取应变率相关特征;同时,根据高分辨率磁共振成像诊断结果,将斑块分为稳定斑块和易损斑块。根据受试者工作特征曲线下的面积对各个特征进行分析,最后选取绝对应变率的99%分位数、最大值、标准差和均值等四个特征,并进行组合,采用径向基函数为核函数的SVM对颈动脉易损斑块进行识别,在测试集上的灵敏性、特异性、准确性分别为70.0%、88.0%、81.3%。本研究初步验证了基于超声弹性成像的SVM在颈动脉易损斑块自动识别中的可行性。
The objective of this study was automatic identification of carotid vulnerable plaques using ultrasound elastography based support vector machine(SVM). Ultrasound radiofrequency data of 80 carotid atherosclerotic plaques from 52 volunteers were acquired in the longitudinal view, and were used to estimate the strain rate distribution with an elastography algorithm.Then the strain rate features of the plaques were extracted. Meanwhile, the plaques were classified to be stable or vulnerable using high-resolution magnetic resonance imaging. The area under the receiver operating characteristic curve was used to analyze each strain rate feature, and the maximum, 99 th percentile, mean, and standard deviation of absolute strain rates were selected and combined. The vulnerable plaques were identified using SVM with radial basis function, achieving sensitivity,specificity, and accuracy of 70.0%, 88.0%, and 81.3%, respectively, in the testing dataset. This study validates the feasibility of ultrasound elastography based SVM in automatic identification of carotid vulnerable plaques.
The objective of this study was automatic identification of carotid vulnerable plaques using ultrasound elastography based support vector machine(SVM). Ultrasound radiofrequency data of 80 carotid atherosclerotic plaques from 52 volunteers were acquired in the longitudinal view, and were used to estimate the strain rate distribution with an elastography algorithm.Then the strain rate features of the plaques were extracted. Meanwhile, the plaques were classified to be stable or vulnerable using high-resolution magnetic resonance imaging. The area under the receiver operating characteristic curve was used to analyze each strain rate feature, and the maximum, 99 th percentile, mean, and standard deviation of absolute strain rates were selected and combined. The vulnerable plaques were identified using SVM with radial basis function, achieving sensitivity,specificity, and accuracy of 70.0%, 88.0%, and 81.3%, respectively, in the testing dataset. This study validates the feasibility of ultrasound elastography based SVM in automatic identification of carotid vulnerable plaques.
引文
[1]陈伟伟,王文,隋辉,等.《中国心血管病报告2016》要点解读[J].中华高血压杂志,2017,32(7):605-608.
[2]Mendel T,Popow J,Hier DB,et al.Advanced atherosclerosis of the aortic arch is uncommon in ischemic stroke:An autopsy study[J].Neurol Res,2002,24(5):491.
[3]Golemati S,Gastounioti A,Nikita KS.Toward novel noninvasive and low-cost markers for predicting strokes in asymptomatic carotid atherosclerosis:The role of ultrasound image analysis[J].IEEE Trans Biom Eng,2013,60(3):652-658.
[4]Streifler JY.Asymptomatic carotid stenosis:intervention or just stick to medical therapy--the case for medical therapy[J].JNeural Trans,2011,118(4):637-640.
[5]Saam T,Underhill HR,Chu B,et al.Prevalence of american heart association type vi carotid atherosclerotic lesions identified by magnetic resonance imaging for different levels of stenosis as measured by duplex ultrasound[J].J Am Coll Cardiol,2008,51(10):1014-1021.
[6]Finn AV,Nakano M,Narula J,et al.Concept of vulnerable/unstable plaque[J].Arterioscler Thromb Vasc Biol,2010,30(7):1282.
[7]Spacek M,Zemanek D,Hutyra M,et al.Vulnerable atherosclerotic plaque-review of current concepts and advanced imaging[D].Czechoslovakia:the Medical Faculty of the University Palacky Olomouc Czechoslovakia,2018.
[8]Nighoghossian N,Derex L,Douek P.The vulnerable carotid artery plaque current imaging methods and new perspectives[J].Stroke,2005,36(12):2764.
[9]Clarke SE,Hammond RR,Mitchell JR,et al.Quantitative assessment of carotid plaque composition using multicontrast MRI and registered histology[J].Magnetic Res Med J,2003,50(6):1199-1208.
[10]Cai JM,Hatsukami TS,Ferguson MS,et al.Classification of human carotid atherosclerotic lesions with in vivo multicontrast magnetic resonance imaging[J].Circulation,2002,106(11):1368-1373.
[11]Ophir J,Garra B,Kallel F,et al.Elastographic imaging[J].Ultrasound Med Biol,2000,26(S1):S23.
[12]de Korte CL,Ignacio Céspedes EI,Af VDS,et al.Intravascular elasticity imaging using ultrasound:feasibility studies in phantoms[J].Ultrasound Med Biol,1997,23(5):735.
[13]Muraki M,Mikami T,Yoshimoto T,et al.Sonographic detection of abnormal plaque motion of the carotid artery:its usefulness in diagnosing high-risk lesions ranging from plaque rupture to ulcer formation[J].Ultrasound Med Biol,2016,42(2):358-364.
[14]Roy Cardinal MH,Mhg H,Qin Z,et al.Carotid artery plaquevul nera bility assess mentusing noninvasive ultrasound elastography:validation with MRI[J].Ajr Am JRoentgenol,2017,209(1):142-151.
[15]Giger ML,Chan HP,Boone J.Anniversary paper:History and status of CAD and quantitative image analysis:The role of Medical Physics and AAPM[J].Med Phys,2008,35(12):5799.
[16]Yu H,Kim S.SVM Tutorial-Classification,Regression and Ranking[J].Handbook Nat Comput,2012,479-506.
[17]张学工.模式识别[M].3版.北京:清华大学出版社,2010.
[18]Vassis D,Kampouraki BA,Belsis P,et al.Using neural networks and SVMs for automatic medical diagnosis:A comprehensive review[A].International Conference on Integrated Information[C].New York:American Institute of Physics,2015.
[19]张麒,汪源源,马剑英,等.基于血管内超声图像自动识别易损斑块[J].光学精密工程,2011,19(10):2507-2519.
[20]Balu N,Yarnykh VL,Chu B,et al.Carotid plaque assessment using fast 3D isotropic-resolution black-blood MRI[J].Magn Reson Med,2011,65(3):627-637.
[21]Fan Z,Zhang Z,Chung YC,et al.Carotid arterial wall MRIat 3T using 3D variable-flip-angle turbo spin-echo(TSE)with flow-sensitive dephasing(FSD)[J].J Magn Reson Imaging,2010,31(3):645.
[22]Wang J,Peter B?rnert,Zhao H,et al.Simultaneous noncontrast angiography and intraplaque hemorrhage(SNAP)imaging for carotid atherosclerotic disease evaluation[J].Magn Reson Med,2013,69(2):337-345.
[23]Zhou Z,Rui L,Zhao X,et al.Evaluation of 3D multi-contrast joint intra-and extracranial vessel wall cardiovascular magnetic resonance[J].J Cardiov Magn Reson,2015,17(1):41.
[24]Pan X,Gao J,Tao S,et al.A two-step optical flow method for strain estimation in elastography:Simulation and phantom study[J].Ultrasonics,2014,54(4):990-996.
[25]Huang C,Pan X,He Q,et al.Ultrasound-based carotid elastography for detection of vulnerable atherosclerotic plaques validated by magnetic resonance imaging[J].Ultrasound Med Biol,2016,42(2):365-377.
[26]Fawcett T.An introduction to ROC analysis[J].Pattern Recogn Lett,2005,27(8):861-874.
[27]Kearns M,Ron D.Algorithmics tability and sanit ycheck bounds for leave-one-out cross-validation[J].Neural Comput,2006,11(6):1427-1453.
[28]Freedman DA.Statistical models:Theory and practice[J].Technometrics,2009,48(2):315.
[29]阮秋琦.数字图像处理学[M].北京:电子工业出版社,2013.
[2]Mendel T,Popow J,Hier DB,et al.Advanced atherosclerosis of the aortic arch is uncommon in ischemic stroke:An autopsy study[J].Neurol Res,2002,24(5):491.
[3]Golemati S,Gastounioti A,Nikita KS.Toward novel noninvasive and low-cost markers for predicting strokes in asymptomatic carotid atherosclerosis:The role of ultrasound image analysis[J].IEEE Trans Biom Eng,2013,60(3):652-658.
[4]Streifler JY.Asymptomatic carotid stenosis:intervention or just stick to medical therapy--the case for medical therapy[J].JNeural Trans,2011,118(4):637-640.
[5]Saam T,Underhill HR,Chu B,et al.Prevalence of american heart association type vi carotid atherosclerotic lesions identified by magnetic resonance imaging for different levels of stenosis as measured by duplex ultrasound[J].J Am Coll Cardiol,2008,51(10):1014-1021.
[6]Finn AV,Nakano M,Narula J,et al.Concept of vulnerable/unstable plaque[J].Arterioscler Thromb Vasc Biol,2010,30(7):1282.
[7]Spacek M,Zemanek D,Hutyra M,et al.Vulnerable atherosclerotic plaque-review of current concepts and advanced imaging[D].Czechoslovakia:the Medical Faculty of the University Palacky Olomouc Czechoslovakia,2018.
[8]Nighoghossian N,Derex L,Douek P.The vulnerable carotid artery plaque current imaging methods and new perspectives[J].Stroke,2005,36(12):2764.
[9]Clarke SE,Hammond RR,Mitchell JR,et al.Quantitative assessment of carotid plaque composition using multicontrast MRI and registered histology[J].Magnetic Res Med J,2003,50(6):1199-1208.
[10]Cai JM,Hatsukami TS,Ferguson MS,et al.Classification of human carotid atherosclerotic lesions with in vivo multicontrast magnetic resonance imaging[J].Circulation,2002,106(11):1368-1373.
[11]Ophir J,Garra B,Kallel F,et al.Elastographic imaging[J].Ultrasound Med Biol,2000,26(S1):S23.
[12]de Korte CL,Ignacio Céspedes EI,Af VDS,et al.Intravascular elasticity imaging using ultrasound:feasibility studies in phantoms[J].Ultrasound Med Biol,1997,23(5):735.
[13]Muraki M,Mikami T,Yoshimoto T,et al.Sonographic detection of abnormal plaque motion of the carotid artery:its usefulness in diagnosing high-risk lesions ranging from plaque rupture to ulcer formation[J].Ultrasound Med Biol,2016,42(2):358-364.
[14]Roy Cardinal MH,Mhg H,Qin Z,et al.Carotid artery plaquevul nera bility assess mentusing noninvasive ultrasound elastography:validation with MRI[J].Ajr Am JRoentgenol,2017,209(1):142-151.
[15]Giger ML,Chan HP,Boone J.Anniversary paper:History and status of CAD and quantitative image analysis:The role of Medical Physics and AAPM[J].Med Phys,2008,35(12):5799.
[16]Yu H,Kim S.SVM Tutorial-Classification,Regression and Ranking[J].Handbook Nat Comput,2012,479-506.
[17]张学工.模式识别[M].3版.北京:清华大学出版社,2010.
[18]Vassis D,Kampouraki BA,Belsis P,et al.Using neural networks and SVMs for automatic medical diagnosis:A comprehensive review[A].International Conference on Integrated Information[C].New York:American Institute of Physics,2015.
[19]张麒,汪源源,马剑英,等.基于血管内超声图像自动识别易损斑块[J].光学精密工程,2011,19(10):2507-2519.
[20]Balu N,Yarnykh VL,Chu B,et al.Carotid plaque assessment using fast 3D isotropic-resolution black-blood MRI[J].Magn Reson Med,2011,65(3):627-637.
[21]Fan Z,Zhang Z,Chung YC,et al.Carotid arterial wall MRIat 3T using 3D variable-flip-angle turbo spin-echo(TSE)with flow-sensitive dephasing(FSD)[J].J Magn Reson Imaging,2010,31(3):645.
[22]Wang J,Peter B?rnert,Zhao H,et al.Simultaneous noncontrast angiography and intraplaque hemorrhage(SNAP)imaging for carotid atherosclerotic disease evaluation[J].Magn Reson Med,2013,69(2):337-345.
[23]Zhou Z,Rui L,Zhao X,et al.Evaluation of 3D multi-contrast joint intra-and extracranial vessel wall cardiovascular magnetic resonance[J].J Cardiov Magn Reson,2015,17(1):41.
[24]Pan X,Gao J,Tao S,et al.A two-step optical flow method for strain estimation in elastography:Simulation and phantom study[J].Ultrasonics,2014,54(4):990-996.
[25]Huang C,Pan X,He Q,et al.Ultrasound-based carotid elastography for detection of vulnerable atherosclerotic plaques validated by magnetic resonance imaging[J].Ultrasound Med Biol,2016,42(2):365-377.
[26]Fawcett T.An introduction to ROC analysis[J].Pattern Recogn Lett,2005,27(8):861-874.
[27]Kearns M,Ron D.Algorithmics tability and sanit ycheck bounds for leave-one-out cross-validation[J].Neural Comput,2006,11(6):1427-1453.
[28]Freedman DA.Statistical models:Theory and practice[J].Technometrics,2009,48(2):315.
[29]阮秋琦.数字图像处理学[M].北京:电子工业出版社,2013.