基于深度卷积网络的阿尔茨海默病诊断模型研究
|
摘要
目的:提出一种基于脑MRI与深度学习和迁移学习准确区分阿尔茨海默病(Alzheimer's disease,AD)与正常老化(normal control,NC)的方法。方法:选取阿尔茨海默病神经影像学组织(Alzheimer Disease Neuroimaging Initiative,ADNI)数据库中194例NC与105例AD受试者的脑结构磁共振成像(structural MRI,sMRI),生成全脑灰质图。基于经典网络AlexNet采用特征迁移学习的方法对AD与NC分别进行特征提取,再结合主成分分析法与序列前向搜索的方法对特征降维与选择,最后运用支持向量机对所选特征进行分类,统计高斯平滑核半高宽(full width at half maximum,FWHM)分别为0、8 mm时在卷积层conv3、conv4、conv5的分类准确率、灵敏度和特异性。结果:在AlexNet第四卷积层(conv4)分类准确率达到最优,在高斯平滑核FWHM为0 mm时,conv4分类准确率为95.14%,灵敏度和特异性分别为96.43%和94.83%。结论:通过该研究提出的分类方法建立的特征迁移学习模型在AD与NC分类中取得较为理想的分类结果,说明该方法是一种可行的分类方法。
Objective To apply computer-aided diagnosis to accurately identify early Alzheimer's disease(AD) and normal control(NC). Methods Totally 194 AD patients from Alzheimer Disease Neuroimaging Initiative(ADNI) and 105 NC subjects underwent structural magnetic resonance imaging(sMRI) to form whole brain gray matter maps. Transfer learning was used for features extraction of AD and NC based on AlexNet, then principal component analysis method was combined with forward sequence selection method to reduce the dimension and select features. Finally, the support vector machine was applied to AD/NC classification and determining the accuracies, sensitivities and specificities of convolutional layer conv3, conv4 and conv5 in case Gaussian kernel full width at half maximum(FWHM) was 0 or 8 mm. Results The classification accuracy gained the highest value on conv4 layer of AlexNet, which reached 95.14% in case Gaussian kernel FWHM was 0 mm, when the sensitivity and specificity were 96.43% and 94.83% respectively. Conclusion The feature transfer learning based on deep learning model proposed in this study proves to behave well in AD and NC classification.
Objective To apply computer-aided diagnosis to accurately identify early Alzheimer's disease(AD) and normal control(NC). Methods Totally 194 AD patients from Alzheimer Disease Neuroimaging Initiative(ADNI) and 105 NC subjects underwent structural magnetic resonance imaging(sMRI) to form whole brain gray matter maps. Transfer learning was used for features extraction of AD and NC based on AlexNet, then principal component analysis method was combined with forward sequence selection method to reduce the dimension and select features. Finally, the support vector machine was applied to AD/NC classification and determining the accuracies, sensitivities and specificities of convolutional layer conv3, conv4 and conv5 in case Gaussian kernel full width at half maximum(FWHM) was 0 or 8 mm. Results The classification accuracy gained the highest value on conv4 layer of AlexNet, which reached 95.14% in case Gaussian kernel FWHM was 0 mm, when the sensitivity and specificity were 96.43% and 94.83% respectively. Conclusion The feature transfer learning based on deep learning model proposed in this study proves to behave well in AD and NC classification.
引文
[1] LIN L,FU Z,XU X,et al. Mouse brain magnetic resonance microscopy:applications in Alzheimer disease[J]. Microsc Res Tech,2015,78(5):416-424.
[2]靳聪,林岚,付振荣,等. MRI在阿尔茨海默氏病中的应用研究进展[J].智慧健康,2015,1(1):52-56.
[3] Alzheimer's Association. 2016 Alzheimer’s disease facts and figures[J]. Alzheimer Dement,2016,12(4):459-509.
[4] ZAREI M,BECKMANN C F,BINNEWIJZEND M A A,et al.Functional segmentation of the hippocampus in the healthy human brain and in Alzheimer’s disease[J]. Neuroimage,2013,66:28-35.
[5] RUAN Q,D'ONOFRIO G,SANCARLO D,et al. Potential neuroimaging biomarkers of pathologic brain changes in mild cognitive impairment and Alzheimer’s disease:a systematic review[J]. BMC Geriatr,2016,16(1):104.
[6] MCKHANN G M. Changing concepts of Alzheimer disease[J].JAMA,2011,305(23):2 458-2 459.
[7] KIM Y K,NA K S. Application of machine learning classification for structural brain MRI in mood disorders:critical review from a clinical perspective[J]. Prog Neuropsychopharmacol Biol Psychiatry,2018,80:71-80.
[8] ADASZEWSKI S,DUKART J,KHERIF F,et al. How early can we predict Alzheimer’s disease using computational anatomy?[J]. Neurobiol Aging,2013,34(12):2 815-2 826.
[9] MORADI E,TOHKA J,GASER C. Semi-supervised learning in MCI-to-AD conversion prediction—when is unlabeled data useful?[C]//Pattern Recognition in Neuroimaging,June 4-6,2014,Tubingen,Germany. New York:IEEE,2014:1-4.
[10] HUANG L,JIN Y,GAO Y,et al. Longitudinal clinical score prediction in Alzheimer’s disease with soft-split sparse regression based random forest[J]. Neurobiol Aging,2016,46:180-191.
[11]田苗,林岚,张柏雯,等.深度学习在神经影像中的应用研究[J].中国医疗设备,2016,31(12):4-9.
[12]张柏雯,林岚,吴水才.深度学习在轻度认知障碍转化与分类中的应用分析[J].医疗卫生装备,2017,38(9):105-111.
[13] SUK H I,SHEN D. Deep learning-based feature representation for AD/MCI classification[J]. Med Image Comput Comput Assist Interv,2013,16(2):583-590.
[14] LIU S,LIU S,CAI W,et al. Multimodal neuroimaging feature learning for multiclass diagnosis of Alzheimer’s disease[J].IEEE Trans Biomed Eng,2015,62(4):1 132-1 140.
[15] SUK H I,LEE S W,SHEN D. Hierarchical feature representation and multimodal fusion with deep learning for AD/MCI diagnosis[J]. Neuroimage,2014,101:569-582.
[16] LITJENS G,KOOI T,BEJNORDI B E,et al. A survey on deep learning in medical image analysis[J]. Med Image Anal,2017,42:60-88.
[17] KRIZHEVSKY A,SUTSKEVER I,HINTON G E. ImageNet classification with deep convolutional neural networks[C]//The 25th International Conference on Neural Information Processing Systems,December 3-6,2012,Lake Tahoe,Nevada.New York:IEEE,2012:1 097-1 105.
[18]刘佳,朱明宇,陈芳,等.基于医学影像智能化分析技术的神经外科精准诊疗研究[J].医疗卫生装备,2018,39(2):1-6.
[19] HOSSEINI-ASL E,KEYNTON R,EL-BAZ A. Alzheimer’s disease diagnostics by adaptation of 3D convolutional network[C]//2016 IEEE International Conference on Image Processing(ICIP),September 25-28,2016,Phoenix,Arizona. New York:IEEE,2016:126-130.
[20] KUNCHEVA L I,FAITHFULL W J. PCA feature extraction for change detection in multidimensional unlabeled data[J].IEEE Trans Neural Netw Learn Syst,2014,25(1):69-80.
[21] HUA J,TEMBE W D,DOUGHERTY E R. Performance of feature-selection methods in the classification of high-dimension data[J]. Pattern Recognit,2009,42(3):409-424.
[22] WEINER M W,VEITCH D P,AISEN P S,et al. The Alzheimer’s Disease Neuroimaging Initiative:a review of papers published since its inception[J]. Alzheimer Dement,2012,8(1):S1-S68.
[23] ASHBURNER J. Computational anatomy with the SPM software[J]. Magn Reson Imaging,2009,27(8):1 163-1 174.
[24] QIAN S,LIU H,LIU C,et al. Adaptive activation functions in convolutional neural networks[J]. Neurocomputing,2018,272:204-212.
[25] JIA Y,SHELHAMER E,DONAHUE J,et al. Caffe:convolutional architecture for fast feature embedding[C]//The 22nd ACM International Conference on Multimedia,November 3-7,2014,Orlando,Florida. New York:ACM,2014:675-678.
[26] CHANG C C,LIN C J. LIBSVM:a library for support vector machines[J]. ACM T Intell Syst Tec,2011,2(3):1-27.
[27] SIMONYAN K,ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[EB/OL].[2018-11-02]. https://arxiv.org/pdf/1409.1556.pdf.
[2]靳聪,林岚,付振荣,等. MRI在阿尔茨海默氏病中的应用研究进展[J].智慧健康,2015,1(1):52-56.
[3] Alzheimer's Association. 2016 Alzheimer’s disease facts and figures[J]. Alzheimer Dement,2016,12(4):459-509.
[4] ZAREI M,BECKMANN C F,BINNEWIJZEND M A A,et al.Functional segmentation of the hippocampus in the healthy human brain and in Alzheimer’s disease[J]. Neuroimage,2013,66:28-35.
[5] RUAN Q,D'ONOFRIO G,SANCARLO D,et al. Potential neuroimaging biomarkers of pathologic brain changes in mild cognitive impairment and Alzheimer’s disease:a systematic review[J]. BMC Geriatr,2016,16(1):104.
[6] MCKHANN G M. Changing concepts of Alzheimer disease[J].JAMA,2011,305(23):2 458-2 459.
[7] KIM Y K,NA K S. Application of machine learning classification for structural brain MRI in mood disorders:critical review from a clinical perspective[J]. Prog Neuropsychopharmacol Biol Psychiatry,2018,80:71-80.
[8] ADASZEWSKI S,DUKART J,KHERIF F,et al. How early can we predict Alzheimer’s disease using computational anatomy?[J]. Neurobiol Aging,2013,34(12):2 815-2 826.
[9] MORADI E,TOHKA J,GASER C. Semi-supervised learning in MCI-to-AD conversion prediction—when is unlabeled data useful?[C]//Pattern Recognition in Neuroimaging,June 4-6,2014,Tubingen,Germany. New York:IEEE,2014:1-4.
[10] HUANG L,JIN Y,GAO Y,et al. Longitudinal clinical score prediction in Alzheimer’s disease with soft-split sparse regression based random forest[J]. Neurobiol Aging,2016,46:180-191.
[11]田苗,林岚,张柏雯,等.深度学习在神经影像中的应用研究[J].中国医疗设备,2016,31(12):4-9.
[12]张柏雯,林岚,吴水才.深度学习在轻度认知障碍转化与分类中的应用分析[J].医疗卫生装备,2017,38(9):105-111.
[13] SUK H I,SHEN D. Deep learning-based feature representation for AD/MCI classification[J]. Med Image Comput Comput Assist Interv,2013,16(2):583-590.
[14] LIU S,LIU S,CAI W,et al. Multimodal neuroimaging feature learning for multiclass diagnosis of Alzheimer’s disease[J].IEEE Trans Biomed Eng,2015,62(4):1 132-1 140.
[15] SUK H I,LEE S W,SHEN D. Hierarchical feature representation and multimodal fusion with deep learning for AD/MCI diagnosis[J]. Neuroimage,2014,101:569-582.
[16] LITJENS G,KOOI T,BEJNORDI B E,et al. A survey on deep learning in medical image analysis[J]. Med Image Anal,2017,42:60-88.
[17] KRIZHEVSKY A,SUTSKEVER I,HINTON G E. ImageNet classification with deep convolutional neural networks[C]//The 25th International Conference on Neural Information Processing Systems,December 3-6,2012,Lake Tahoe,Nevada.New York:IEEE,2012:1 097-1 105.
[18]刘佳,朱明宇,陈芳,等.基于医学影像智能化分析技术的神经外科精准诊疗研究[J].医疗卫生装备,2018,39(2):1-6.
[19] HOSSEINI-ASL E,KEYNTON R,EL-BAZ A. Alzheimer’s disease diagnostics by adaptation of 3D convolutional network[C]//2016 IEEE International Conference on Image Processing(ICIP),September 25-28,2016,Phoenix,Arizona. New York:IEEE,2016:126-130.
[20] KUNCHEVA L I,FAITHFULL W J. PCA feature extraction for change detection in multidimensional unlabeled data[J].IEEE Trans Neural Netw Learn Syst,2014,25(1):69-80.
[21] HUA J,TEMBE W D,DOUGHERTY E R. Performance of feature-selection methods in the classification of high-dimension data[J]. Pattern Recognit,2009,42(3):409-424.
[22] WEINER M W,VEITCH D P,AISEN P S,et al. The Alzheimer’s Disease Neuroimaging Initiative:a review of papers published since its inception[J]. Alzheimer Dement,2012,8(1):S1-S68.
[23] ASHBURNER J. Computational anatomy with the SPM software[J]. Magn Reson Imaging,2009,27(8):1 163-1 174.
[24] QIAN S,LIU H,LIU C,et al. Adaptive activation functions in convolutional neural networks[J]. Neurocomputing,2018,272:204-212.
[25] JIA Y,SHELHAMER E,DONAHUE J,et al. Caffe:convolutional architecture for fast feature embedding[C]//The 22nd ACM International Conference on Multimedia,November 3-7,2014,Orlando,Florida. New York:ACM,2014:675-678.
[26] CHANG C C,LIN C J. LIBSVM:a library for support vector machines[J]. ACM T Intell Syst Tec,2011,2(3):1-27.
[27] SIMONYAN K,ZISSERMAN A. Very deep convolutional networks for large-scale image recognition[EB/OL].[2018-11-02]. https://arxiv.org/pdf/1409.1556.pdf.