KPCA和Adaboost算法在阿尔茨海默症功能磁共振影像分类中的应用
|
摘要
本研究的目的在于使用机器学习方法,对脑部功能磁共振成像数据进行分析与特征提取,完成对阿尔茨海默症(AD)的辅助诊断与分析。首先对数据进行预处理与去除协变量,并从大脑全局特征出发,根据现有的自动解剖标记模板,把每个被试的大脑分为116个脑区,通过提取每个脑区的时间序列,构建全脑功能连接矩阵,然后使用核主成分分析法进行特征提取,最后用Adaboost算法进行分类。在对34名AD患者、35名轻度认知障碍患者和35名正常对照组的功能磁共振成像数据进行的实验结果表明,利用静息态功能磁共振成像,同时结合机器学习的方法,能够有效地实现AD的正确分类,准确率可以达到96%,该结果可以为AD患者的临床辅助诊断提供有效的判断依据。
The purpose of this study is to achieve the auxiliary diagnosis and analysis of Alzheimer's disease(AD) by analyzing and characterizing brain functional magnetic resonance imaging(f MRI) data using machine learning method. After the f MRI data is preprocessed and the covariate is removed, the brain of each subject is divided into 116 brain regions according to anatomical automatic labeling template, and the whole brain functional connection matrix is constructed by extracting the time series of each brain region. Kernel principal component analysis is used to extract features and Adaboost algorithm is used for classification. The results of the experiment on f MRI images of 34 patients with AD, 35 patients with mild cognitive impairments and 35 normal controls show that using resting state f MRI combined with machine learning method can effectively realize the accurate classification of AD, with a classification accuracy rate up to 96%. The proposed method can provide an effective basis for the auxiliary diagnosis of patients with AD.
The purpose of this study is to achieve the auxiliary diagnosis and analysis of Alzheimer's disease(AD) by analyzing and characterizing brain functional magnetic resonance imaging(f MRI) data using machine learning method. After the f MRI data is preprocessed and the covariate is removed, the brain of each subject is divided into 116 brain regions according to anatomical automatic labeling template, and the whole brain functional connection matrix is constructed by extracting the time series of each brain region. Kernel principal component analysis is used to extract features and Adaboost algorithm is used for classification. The results of the experiment on f MRI images of 34 patients with AD, 35 patients with mild cognitive impairments and 35 normal controls show that using resting state f MRI combined with machine learning method can effectively realize the accurate classification of AD, with a classification accuracy rate up to 96%. The proposed method can provide an effective basis for the auxiliary diagnosis of patients with AD.
引文
[1] ZHANG J, GAO Y, GAO Y Z, et al. Detecting anatomical landmarks for fast Alzheimer's disease diagnosis[J]. IEEE Trans Med Imaging,2016, 35(12):2524-2533.
[2]樊东琼,李锐,雷旭,等.阿尔兹海默症及轻度认知障碍静息态大尺度脑网络功能连接的改变[J].心理科学进展, 2016, 24(2):217-227.FAN D Q, LI R, LEI X, et al. Changes in functional connectivity of large-scale brain networks in resting state of Alzheimer's disease and mild cognitive impairment[J]. Advances in Psychological Science, 2016, 24(2):217-227.
[3] RATHORE S, HABES M, AKSAM I M, et al. A review on neuroimaging-based classification studies and associated feature extraction methods for Alzheimer's disease and its prodromal stages[J]. Neuroimage, 2017, 155:530.
[4]周文,王瑜,肖红兵,等.基于KPCA算法的阿尔茨海默症辅助诊断[J].中国医学物理学杂志, 2018, 35(4):404-409.ZHOU W, WANG Y, XIAO H B, et al. Assisted diagnosis of Alzheimer's disease based on KPCA algorithm[J]. Chinese Journal of Medical Physics, 2018, 35(4):404-409.
[5] KHAZAEE A, EBRAHIMZADEH A, BABAJANIFEREMI A.Classification of patients with MCI and AD from healthy controls using directed graph measures of resting-state f MRI[J]. Behav Brain Res, 2017, 322(Pt B):339-350.
[6] FORD J, FARID H, MAKEDON F, et al. Patient classification of f MRI activation maps[C]//Medical Image Computing and ComputerAssisted Intervention. DBLP, 2003:58-65.
[7] GUO C C, TAN R, HODGES J R, et al. Network-selective vulnerability of the human cerebellum to Alzheimer's disease and frontotemporal dementia[J]. Brain, 2016, 139(Pt 5):1527-1538.
[8] WEINER M W, VEITCH D P, AISEN P S, et al. Recent publications from the Alzheimer's disease neuroimaging initiative:reviewing progress toward improved AD clinical trials[J]. Alzheimers Dement,2017, 13(4):e1.
[9] SONG X W, DONG Z Y, LONG X Y, et al. REST:a toolkit for restingstate functional magnetic resonance imaging data processing[J]. PLo S One, 2011, 6(9):e25031.
[10]KWAK Y, PELTIER S J, BOHNEN N I, et al. L-DOPA changes spontaneous low-frequency BOLD signal oscillations in Parkinson's disease:a resting state fMRI study[J]. Front Syst Neurosci, 2012, 6(6):52.
[11]李亚鹏,覃媛媛,李炜.阿尔茨海默病患者大脑功能网络的改变[J].中国医学物理学杂志, 2013, 30(6):4510-4514.LI Y P, QIN Y Y, LI W. The functional brain network changes of Alzheimer's disease[J]. Chinese Journal of Medical Physics, 2013, 30(6):4510-4514.
[12]ZHANG M, XIA Z F, ZHANG F, et al. Cognitive effects of highfrequency repetitive transcranial magnetic stimulation in Alzheimer's disease:a pilot clinical study[J]. Alzheimers Dement, 2017, 13(7):P260-P261.
[13]刘美洁.脑磁共振成像数据的多类模式分析[D].长沙:国防科学技术大学, 2011.LIU M J. Multi-class pattern analysis on human brain MRI dataset[D]. Changsha:National University of Defense Technology, 2011.
[14]PEARSON K. On lines and planes of closest fit to systems of points in space[J]. Philos Mag, 1901, 2(6):559-572.
[15]SCH?LKOPF B, SMOLA A, MüLLER K R. Nonlinear component analysis as a kernel eigenvalue problem[J]. Neural Comput, 1996, 10(5):1299-1319.
[16]ZHANG J, TUO X, YUAN Z, et al. Analysis of fMRI data using an integrated principal component analysis and supervised affinity propagation clustering approach[J]. IEEE Trans Biomed Eng, 2011,58(11):3184-3196.
[17]CORTES C, VAPNIK V. Support-vector networks[J]. Mach Learn,1995, 20(3):273-297.
[18]WANG Z, CHILDRESS A R, WANG J, et al. Support vector machine learning-based fMRI data group analysis[J]. Neuroimage, 2007, 36(4):1139-1151.
[19]ZHU J, ZOU H, ROSSET S, et al. Multi-class AdaBoost[J]. Stat Interface, 2009, 2(3):349-360.
[20]PEDREGOSA F, GRAMFORT A, MICHEL V, et al. Scikit-learn:machine learning in python[J]. J Mach Learn Res, 2013, 12(10):2825-2830.
[2]樊东琼,李锐,雷旭,等.阿尔兹海默症及轻度认知障碍静息态大尺度脑网络功能连接的改变[J].心理科学进展, 2016, 24(2):217-227.FAN D Q, LI R, LEI X, et al. Changes in functional connectivity of large-scale brain networks in resting state of Alzheimer's disease and mild cognitive impairment[J]. Advances in Psychological Science, 2016, 24(2):217-227.
[3] RATHORE S, HABES M, AKSAM I M, et al. A review on neuroimaging-based classification studies and associated feature extraction methods for Alzheimer's disease and its prodromal stages[J]. Neuroimage, 2017, 155:530.
[4]周文,王瑜,肖红兵,等.基于KPCA算法的阿尔茨海默症辅助诊断[J].中国医学物理学杂志, 2018, 35(4):404-409.ZHOU W, WANG Y, XIAO H B, et al. Assisted diagnosis of Alzheimer's disease based on KPCA algorithm[J]. Chinese Journal of Medical Physics, 2018, 35(4):404-409.
[5] KHAZAEE A, EBRAHIMZADEH A, BABAJANIFEREMI A.Classification of patients with MCI and AD from healthy controls using directed graph measures of resting-state f MRI[J]. Behav Brain Res, 2017, 322(Pt B):339-350.
[6] FORD J, FARID H, MAKEDON F, et al. Patient classification of f MRI activation maps[C]//Medical Image Computing and ComputerAssisted Intervention. DBLP, 2003:58-65.
[7] GUO C C, TAN R, HODGES J R, et al. Network-selective vulnerability of the human cerebellum to Alzheimer's disease and frontotemporal dementia[J]. Brain, 2016, 139(Pt 5):1527-1538.
[8] WEINER M W, VEITCH D P, AISEN P S, et al. Recent publications from the Alzheimer's disease neuroimaging initiative:reviewing progress toward improved AD clinical trials[J]. Alzheimers Dement,2017, 13(4):e1.
[9] SONG X W, DONG Z Y, LONG X Y, et al. REST:a toolkit for restingstate functional magnetic resonance imaging data processing[J]. PLo S One, 2011, 6(9):e25031.
[10]KWAK Y, PELTIER S J, BOHNEN N I, et al. L-DOPA changes spontaneous low-frequency BOLD signal oscillations in Parkinson's disease:a resting state fMRI study[J]. Front Syst Neurosci, 2012, 6(6):52.
[11]李亚鹏,覃媛媛,李炜.阿尔茨海默病患者大脑功能网络的改变[J].中国医学物理学杂志, 2013, 30(6):4510-4514.LI Y P, QIN Y Y, LI W. The functional brain network changes of Alzheimer's disease[J]. Chinese Journal of Medical Physics, 2013, 30(6):4510-4514.
[12]ZHANG M, XIA Z F, ZHANG F, et al. Cognitive effects of highfrequency repetitive transcranial magnetic stimulation in Alzheimer's disease:a pilot clinical study[J]. Alzheimers Dement, 2017, 13(7):P260-P261.
[13]刘美洁.脑磁共振成像数据的多类模式分析[D].长沙:国防科学技术大学, 2011.LIU M J. Multi-class pattern analysis on human brain MRI dataset[D]. Changsha:National University of Defense Technology, 2011.
[14]PEARSON K. On lines and planes of closest fit to systems of points in space[J]. Philos Mag, 1901, 2(6):559-572.
[15]SCH?LKOPF B, SMOLA A, MüLLER K R. Nonlinear component analysis as a kernel eigenvalue problem[J]. Neural Comput, 1996, 10(5):1299-1319.
[16]ZHANG J, TUO X, YUAN Z, et al. Analysis of fMRI data using an integrated principal component analysis and supervised affinity propagation clustering approach[J]. IEEE Trans Biomed Eng, 2011,58(11):3184-3196.
[17]CORTES C, VAPNIK V. Support-vector networks[J]. Mach Learn,1995, 20(3):273-297.
[18]WANG Z, CHILDRESS A R, WANG J, et al. Support vector machine learning-based fMRI data group analysis[J]. Neuroimage, 2007, 36(4):1139-1151.
[19]ZHU J, ZOU H, ROSSET S, et al. Multi-class AdaBoost[J]. Stat Interface, 2009, 2(3):349-360.
[20]PEDREGOSA F, GRAMFORT A, MICHEL V, et al. Scikit-learn:machine learning in python[J]. J Mach Learn Res, 2013, 12(10):2825-2830.