RESEARCH ARTICLE
Independent Component Analysis-Based Identification of Covariance Patterns of Microstructural White Matter Damage in Alzheimer’s Disease Xin Ouyang1, Kewei Chen2, Li Yao1,3, Xia Wu1,3, Jiacai Zhang1, Ke Li4, Zhen Jin4, Xiaojuan Guo1,3*, for the Alzheimer’s Disease Neuroimaging Initiative¶
a11111
1 College of Information Science and Technology, Beijing Normal University, Beijing, China, 2 Banner Alzheimer’s Institute and Banner Good Samaritan PET Center, Phoenix, Arizona, United States of America, 3 State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, China, 4 Laboratory of Magnetic Resonance Imaging, Beijing 306 Hospital, Beijing, China ¶ Membership of the Alzheimer’s Disease Neuroimaging Initiative is listed in the Acknowledgments. * [email protected]
OPEN ACCESS Citation: Ouyang X, Chen K, Yao L, Wu X, Zhang J, Li K, et al. (2015) Independent Component AnalysisBased Identification of Covariance Patterns of Microstructural White Matter Damage in Alzheimer’s Disease. PLoS ONE 10(3): e0119714. doi:10.1371/ journal.pone.0119714 Academic Editor: Yong Liu, Brainnetome Center & National Laboratory of Pattern Recognition, CHINA Received: July 10, 2014 Accepted: January 16, 2015 Published: March 16, 2015 Copyright: © 2015 Ouyang et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All relevant data are within the paper and its Supporting Information files. Funding: This work was supported by the National Key Basic Research Program (973 Program), China (2012CB720704), the National Natural Science Foundation (NNSF), China (81000603), the Funds for International Cooperation and Exchange of NNSF, China (61210001), Key Program of NNSF, China (91320201), the National Institute of Mental Health, US (RO1 MH57899), the National Institute on Aging, US (9R01AG031581-10, P30 AG19610, k23 AG24062), and the State of Arizona. Data collection
Abstract The existing DTI studies have suggested that white matter damage constitutes an important part of the neurodegenerative changes in Alzheimer’s disease (AD). The present study aimed to identify the regional covariance patterns of microstructural white matter changes associated with AD. In this study, we applied a multivariate analysis approach, independent component analysis (ICA), to identify covariance patterns of microstructural white matter damage based on fractional anisotropy (FA) skeletonised images from DTI data in 39 AD patients and 41 healthy controls (HCs) from the Alzheimer’s Disease Neuroimaging Initiative database. The multivariate ICA decomposed the subject-dimension concatenated FA data into a mixing coefficient matrix and a source matrix. Twenty-eight independent components (ICs) were extracted, and a two sample t-test on each column of the corresponding mixing coefficient matrix revealed significant AD/HC differences in ICA weights for 7 ICs. The covariant FA changes primarily involved the bilateral corona radiata, the superior longitudinal fasciculus, the cingulum, the hippocampal commissure, and the corpus callosum in AD patients compared to HCs. Our findings identified covariant white matter damage associated with AD based on DTI in combination with multivariate ICA, potentially expanding our understanding of the neuropathological mechanisms of AD.
Introduction Alzheimer’s disease (AD) is a common progressive neurodegenerative disease among the elderly. The prominent morphological alterations in AD involve not only grey matter atrophy but also white matter damage; for review, see [1,2]. DTI, an advanced non-invasive MRI technique, can quantitatively detect the diffusion characteristics that represent the microstructure of white
PLOS ONE | DOI:10.1371/journal.pone.0119714 March 16, 2015
1 / 12
Covariance Patterns of White Matter in AD
and sharing for this project was funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; BioClinica, Inc.; Biogen Idec Inc.; BristolMyers Squibb Company; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; GE Healthcare; Innogenetics, N.V.; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Medpace, Inc.; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Synarc Inc.; and Takeda Pharmaceutical Company. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Disease Cooperative Study at the University of California, San Diego. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: Co-authors Kewei Chen and Xia Wu are PLOS ONE Editorial Board members. This does not alter the authors’ adherence to PLOS ONE Editorial policies and criteria. The data for this project is from a commercial source (ADNI). But this does not alter the authors’ adherence to PLOS ONE policies on sharing data and materials.
matter in the brain in vivo. The existing DTI studies have suggested that white matter damage constitutes an important part of the neurodegenerative changes associated with AD [3–5]. Multiple DTI indices, such as fractional anisotropy (FA), radial diffusivity (RD), axial diffusivity (AD) and mean diffusivity (MD), can detect the abnormalities in white matter in AD [6–10]. Among these indices, FA, which represents the degree of anisotropy of water diffusion, is one of the most common and important parameters used to characterise the microstructural characteristics of white matter [3,8,11,12]. Earlier DTI studies applied regions of interest (ROI)-based method [13–15]. However, the definition of ROIs requires a priori hypothesis, and this approach is subject to the partial volume effect, especially with respect to small or thin white fibre tracts [3,8]. More recently, some DTI studies performed automated voxel-based style analysis across the entire brain [3,11,16], although misregistration and smoothness-related problems may yield unreliable findings. Then Smith et al. proposed a tract-based spatial statistics (TBSS) method, which utilises intermediate non-linear registration and projection onto the mean FA skeleton image to improve the sensitivity, objectivity and interpretability of voxel-wise analysis of multi-subject DTI studies [17,18]. The above DTI studies applied univariate measures and solely focused on the differences between two groups. However, multivariate approaches, such as independent component analysis (ICA), can elucidate the covariance information hidden in imaging data, revealing the relationships between morphological characteristics and providing a representation of the identified network patterns [19,20]. An ICA model is based on a basic assumption that the unobservable source signals are statistical independent and non-Gaussian. The observed or measured signals, such as brain imaging data, can be considered as linear mixtures of independent components. Therefore, ICA is an unsupervised data-driven statistical method that decomposes linearly mixed signals into maximally independent components carrying similar inter-subject covariance information. Moreover, the mixing coefficients can be used to perform statistical analysis to examine the between-group difference. Determining the alterations in the microstructural connections of white matter associated with AD could play a vital role in understanding the fundamental pathology of AD. The purpose of the present study was to identify the regional covariance patterns of microstructural white matter changes associated with AD. We performed multivariate ICA on skeletonised FA images from patients with AD and healthy controls (HCs). Moreover, we assessed the discrimination ability between the AD and HC groups using the subject-specific mixing coefficients for each independent component.
Materials and Methods Data used in the preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). The ADNI was launched in 2003 by the National Institute on Aging (NIA), the National Institute of Biomedical Imaging and Bioengineering (NIBIB), the Food and Drug Administration (FDA), private pharmaceutical companies and non-profit organizations, as a $60 million, 5-year public-private partnership. The primary goal of ADNI has been to test whether serial magnetic resonance imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment (MCI) and early Alzheimer’s disease (AD). Determination of sensitive and specific markers of very early AD progression is intended to aid researchers and clinicians to develop new treatments and monitor their effectiveness, as well as lessen the time and cost of clinical trials. The Principal Investigator of this initiative is Michael W. Weiner, MD, VA Medical Center and University of California—San Francisco. ADNI is the result of efforts of many co-
PLOS ONE | DOI:10.1371/journal.pone.0119714 March 16, 2015
2 / 12
Covariance Patterns of White Matter in AD
investigators from a broad range of academic institutions and private corporations, and subjects have been recruited from over 50 sites across the U.S. and Canada. The initial goal of ADNI was to recruit 800 subjects but ADNI has been followed by ADNI-GO and ADNI-2. To date these three protocols have recruited over 1500 adults, ages 55 to 90, to participate in the research, consisting of cognitively normal older individuals, people with early or late MCI, and people with early AD. The follow up duration of each group is specified in the protocols for ADNI-1, ADNI-2 and ADNI-GO. Subjects originally recruited for ADNI-1 and ADNI-GO had the option to be followed in ADNI-2. For up-to-date information, see www.adni-info.org/.
Ethics statement The Alzheimer’s Disease Neuroimaging Initiative (ADNI) study was approved by Institutional Review Board (IRB) of each participating site including University of Southern California and Banner Alzheimer’s Institute and was conducted in accordance with federal regulations and the Internal Conference on Harmonisation (ICH) guidelines of Good Clinical Practice (GCP). The study subjects provided their written informed consent at the time of enrolment regarding imaging data and completed questionnaires approved by each participating site’s IRB.
Subjects According to the ADNI protocol, the diagnosis of probable AD met the National Institute of Neurological and Communicative Disorders and Stroke/Alzheimer’s Disease and Related Disorders Association (NINCDS/ADRDA) criteria, and the severity of cognitive impairment was determined based on the Mini-Mental State Examination (MMSE) and Clinical Dementia Rating (CDR) scores. In this investigation, the participants included 39 AD patients (23 males and 16 females, mean age: 74.91±8.13 years, range: 60–90 years; mean Mini-Mental State Examination (MMSE) score: 22.87±2.32, range: 18–27; CDR score: 0.5 or 1) and 41 HCs (20 males and 21 females, mean age: 73.97±6.34 years, range: 60–90 years; MMSE score: 29.07±0.96, range: 27–30; CDR: 0). The AD group did not significantly differ from the HC group with respect to gender ratio (w2ð1Þ ¼ 0:836; p ¼ 0:361) or age (t(78) = 0.575,p = 0.567) but exhibited significantly lower MMSE scores (t(78) = -15.768,p = 5.693e-026).
DTI data acquisition All DTI scans were acquired using 3T GE MEDICAL SYSTEM scanners at the various ADNI sites. All scans were collected using the standard ADNI MRI protocol. For each subject, the scans were collected using the following parameters: pulse sequence = EP/SE; matrix size = 256 mm × 256 mm; flip angle = 90°; slice thickness = 2.7 mm; gradient directions = 41 (b = 1000 s/mm2) and 5 acquisitions without diffusion weighting (b = 0 s/mm2).
DTI data pre-processing The DTI data were pre-processed using the FMRIB Software Library (FSL) version 5.0 (http:// www.fmrib.ox.ac.uk/fsl). Using FMRIB’s Diffusion Toolbox (FDT) version 2.0, the eddy current correction and the head motion correction were applied via affine registration of each subject’s diffusion-weighted image to the non-diffusion-weighted image. The non-brain structures were removed using the Brain Extraction Tool, and the FA maps were generated based on the diffusion tensor reconstructed using the DTIfit program. Then, all subjects’ FA images were processed using TBSS analysis [17,18]: first, each subject’s FA image was nonlinear registrated to the MNI space; second, the mean FA image was calculated and thinned to generate the mean FA skeleton image (FA > 0.2), which represents the centre of white matter tract; third,
PLOS ONE | DOI:10.1371/journal.pone.0119714 March 16, 2015
3 / 12
Covariance Patterns of White Matter in AD
each subject’s transformed FA image was projected to the mean FA skeleton image by calculating the maximum FA values from the nearest tract centre and filling the corresponding position in the skeleton. Lastly, all subjects’ skeletonised FA images were utilised for multivariate ICA.
Multivariate ICA ICA was performed using the Fusion ICA toolbox (FIT) version 2.0c (http://icatb.sourceforge. net/). FIT toolbox is used to examine the shared information between the features and it can also be used to analyze one modality data. In the current study, we did not apply fusion theory and just analyzed DTI data using single modality ICA. The ICA model can be written as X ¼ AS; where X is an observed data matrix, A is an unknown mixing matrix and S is a source matrix that refers to the statistically independent source signals. The ICA model assumes that X is generated by linearly mixing A and S. In this study, the skeletonised FA image of each subject was arrayed into a one-dimensional vector, and the input data matrix X (size: subject number by voxel number) was generated by concatenating the skeletonised FA data (a row vector) of all subjects from the two groups, AD patients and HCs. The initial data matrix X was decomposed using ICA based on the Infomax algorithm to obtain the mixing coefficient matrix A (size: subject number by source number) and the source matrix S (size: source number by voxel number) [21]. Each column of the mixing matrix A (ICA weights) indicates the degree to which each subject expresses the source or network differences between two groups. Each row of the source matrix S represents the source maps exhibiting the inter-subject covariance information.
Statistical analysis A two-sample t-test was performed on each column of the mixing coefficient matrix (ICA weights) to examine the between-group differences (p
Independent Component Analysis-Based Identification of Covariance Patterns of Microstructural White Matter Damage in Alzheimer’s Disease Xin Ouyang1, Kewei Chen2, Li Yao1,3, Xia Wu1,3, Jiacai Zhang1, Ke Li4, Zhen Jin4, Xiaojuan Guo1,3*, for the Alzheimer’s Disease Neuroimaging Initiative¶
a11111
1 College of Information Science and Technology, Beijing Normal University, Beijing, China, 2 Banner Alzheimer’s Institute and Banner Good Samaritan PET Center, Phoenix, Arizona, United States of America, 3 State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, China, 4 Laboratory of Magnetic Resonance Imaging, Beijing 306 Hospital, Beijing, China ¶ Membership of the Alzheimer’s Disease Neuroimaging Initiative is listed in the Acknowledgments. * [email protected]
OPEN ACCESS Citation: Ouyang X, Chen K, Yao L, Wu X, Zhang J, Li K, et al. (2015) Independent Component AnalysisBased Identification of Covariance Patterns of Microstructural White Matter Damage in Alzheimer’s Disease. PLoS ONE 10(3): e0119714. doi:10.1371/ journal.pone.0119714 Academic Editor: Yong Liu, Brainnetome Center & National Laboratory of Pattern Recognition, CHINA Received: July 10, 2014 Accepted: January 16, 2015 Published: March 16, 2015 Copyright: © 2015 Ouyang et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All relevant data are within the paper and its Supporting Information files. Funding: This work was supported by the National Key Basic Research Program (973 Program), China (2012CB720704), the National Natural Science Foundation (NNSF), China (81000603), the Funds for International Cooperation and Exchange of NNSF, China (61210001), Key Program of NNSF, China (91320201), the National Institute of Mental Health, US (RO1 MH57899), the National Institute on Aging, US (9R01AG031581-10, P30 AG19610, k23 AG24062), and the State of Arizona. Data collection
Abstract The existing DTI studies have suggested that white matter damage constitutes an important part of the neurodegenerative changes in Alzheimer’s disease (AD). The present study aimed to identify the regional covariance patterns of microstructural white matter changes associated with AD. In this study, we applied a multivariate analysis approach, independent component analysis (ICA), to identify covariance patterns of microstructural white matter damage based on fractional anisotropy (FA) skeletonised images from DTI data in 39 AD patients and 41 healthy controls (HCs) from the Alzheimer’s Disease Neuroimaging Initiative database. The multivariate ICA decomposed the subject-dimension concatenated FA data into a mixing coefficient matrix and a source matrix. Twenty-eight independent components (ICs) were extracted, and a two sample t-test on each column of the corresponding mixing coefficient matrix revealed significant AD/HC differences in ICA weights for 7 ICs. The covariant FA changes primarily involved the bilateral corona radiata, the superior longitudinal fasciculus, the cingulum, the hippocampal commissure, and the corpus callosum in AD patients compared to HCs. Our findings identified covariant white matter damage associated with AD based on DTI in combination with multivariate ICA, potentially expanding our understanding of the neuropathological mechanisms of AD.
Introduction Alzheimer’s disease (AD) is a common progressive neurodegenerative disease among the elderly. The prominent morphological alterations in AD involve not only grey matter atrophy but also white matter damage; for review, see [1,2]. DTI, an advanced non-invasive MRI technique, can quantitatively detect the diffusion characteristics that represent the microstructure of white
PLOS ONE | DOI:10.1371/journal.pone.0119714 March 16, 2015
1 / 12
Covariance Patterns of White Matter in AD
and sharing for this project was funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; BioClinica, Inc.; Biogen Idec Inc.; BristolMyers Squibb Company; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; GE Healthcare; Innogenetics, N.V.; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Medpace, Inc.; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Synarc Inc.; and Takeda Pharmaceutical Company. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Disease Cooperative Study at the University of California, San Diego. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: Co-authors Kewei Chen and Xia Wu are PLOS ONE Editorial Board members. This does not alter the authors’ adherence to PLOS ONE Editorial policies and criteria. The data for this project is from a commercial source (ADNI). But this does not alter the authors’ adherence to PLOS ONE policies on sharing data and materials.
matter in the brain in vivo. The existing DTI studies have suggested that white matter damage constitutes an important part of the neurodegenerative changes associated with AD [3–5]. Multiple DTI indices, such as fractional anisotropy (FA), radial diffusivity (RD), axial diffusivity (AD) and mean diffusivity (MD), can detect the abnormalities in white matter in AD [6–10]. Among these indices, FA, which represents the degree of anisotropy of water diffusion, is one of the most common and important parameters used to characterise the microstructural characteristics of white matter [3,8,11,12]. Earlier DTI studies applied regions of interest (ROI)-based method [13–15]. However, the definition of ROIs requires a priori hypothesis, and this approach is subject to the partial volume effect, especially with respect to small or thin white fibre tracts [3,8]. More recently, some DTI studies performed automated voxel-based style analysis across the entire brain [3,11,16], although misregistration and smoothness-related problems may yield unreliable findings. Then Smith et al. proposed a tract-based spatial statistics (TBSS) method, which utilises intermediate non-linear registration and projection onto the mean FA skeleton image to improve the sensitivity, objectivity and interpretability of voxel-wise analysis of multi-subject DTI studies [17,18]. The above DTI studies applied univariate measures and solely focused on the differences between two groups. However, multivariate approaches, such as independent component analysis (ICA), can elucidate the covariance information hidden in imaging data, revealing the relationships between morphological characteristics and providing a representation of the identified network patterns [19,20]. An ICA model is based on a basic assumption that the unobservable source signals are statistical independent and non-Gaussian. The observed or measured signals, such as brain imaging data, can be considered as linear mixtures of independent components. Therefore, ICA is an unsupervised data-driven statistical method that decomposes linearly mixed signals into maximally independent components carrying similar inter-subject covariance information. Moreover, the mixing coefficients can be used to perform statistical analysis to examine the between-group difference. Determining the alterations in the microstructural connections of white matter associated with AD could play a vital role in understanding the fundamental pathology of AD. The purpose of the present study was to identify the regional covariance patterns of microstructural white matter changes associated with AD. We performed multivariate ICA on skeletonised FA images from patients with AD and healthy controls (HCs). Moreover, we assessed the discrimination ability between the AD and HC groups using the subject-specific mixing coefficients for each independent component.
Materials and Methods Data used in the preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). The ADNI was launched in 2003 by the National Institute on Aging (NIA), the National Institute of Biomedical Imaging and Bioengineering (NIBIB), the Food and Drug Administration (FDA), private pharmaceutical companies and non-profit organizations, as a $60 million, 5-year public-private partnership. The primary goal of ADNI has been to test whether serial magnetic resonance imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment (MCI) and early Alzheimer’s disease (AD). Determination of sensitive and specific markers of very early AD progression is intended to aid researchers and clinicians to develop new treatments and monitor their effectiveness, as well as lessen the time and cost of clinical trials. The Principal Investigator of this initiative is Michael W. Weiner, MD, VA Medical Center and University of California—San Francisco. ADNI is the result of efforts of many co-
PLOS ONE | DOI:10.1371/journal.pone.0119714 March 16, 2015
2 / 12
Covariance Patterns of White Matter in AD
investigators from a broad range of academic institutions and private corporations, and subjects have been recruited from over 50 sites across the U.S. and Canada. The initial goal of ADNI was to recruit 800 subjects but ADNI has been followed by ADNI-GO and ADNI-2. To date these three protocols have recruited over 1500 adults, ages 55 to 90, to participate in the research, consisting of cognitively normal older individuals, people with early or late MCI, and people with early AD. The follow up duration of each group is specified in the protocols for ADNI-1, ADNI-2 and ADNI-GO. Subjects originally recruited for ADNI-1 and ADNI-GO had the option to be followed in ADNI-2. For up-to-date information, see www.adni-info.org/.
Ethics statement The Alzheimer’s Disease Neuroimaging Initiative (ADNI) study was approved by Institutional Review Board (IRB) of each participating site including University of Southern California and Banner Alzheimer’s Institute and was conducted in accordance with federal regulations and the Internal Conference on Harmonisation (ICH) guidelines of Good Clinical Practice (GCP). The study subjects provided their written informed consent at the time of enrolment regarding imaging data and completed questionnaires approved by each participating site’s IRB.
Subjects According to the ADNI protocol, the diagnosis of probable AD met the National Institute of Neurological and Communicative Disorders and Stroke/Alzheimer’s Disease and Related Disorders Association (NINCDS/ADRDA) criteria, and the severity of cognitive impairment was determined based on the Mini-Mental State Examination (MMSE) and Clinical Dementia Rating (CDR) scores. In this investigation, the participants included 39 AD patients (23 males and 16 females, mean age: 74.91±8.13 years, range: 60–90 years; mean Mini-Mental State Examination (MMSE) score: 22.87±2.32, range: 18–27; CDR score: 0.5 or 1) and 41 HCs (20 males and 21 females, mean age: 73.97±6.34 years, range: 60–90 years; MMSE score: 29.07±0.96, range: 27–30; CDR: 0). The AD group did not significantly differ from the HC group with respect to gender ratio (w2ð1Þ ¼ 0:836; p ¼ 0:361) or age (t(78) = 0.575,p = 0.567) but exhibited significantly lower MMSE scores (t(78) = -15.768,p = 5.693e-026).
DTI data acquisition All DTI scans were acquired using 3T GE MEDICAL SYSTEM scanners at the various ADNI sites. All scans were collected using the standard ADNI MRI protocol. For each subject, the scans were collected using the following parameters: pulse sequence = EP/SE; matrix size = 256 mm × 256 mm; flip angle = 90°; slice thickness = 2.7 mm; gradient directions = 41 (b = 1000 s/mm2) and 5 acquisitions without diffusion weighting (b = 0 s/mm2).
DTI data pre-processing The DTI data were pre-processed using the FMRIB Software Library (FSL) version 5.0 (http:// www.fmrib.ox.ac.uk/fsl). Using FMRIB’s Diffusion Toolbox (FDT) version 2.0, the eddy current correction and the head motion correction were applied via affine registration of each subject’s diffusion-weighted image to the non-diffusion-weighted image. The non-brain structures were removed using the Brain Extraction Tool, and the FA maps were generated based on the diffusion tensor reconstructed using the DTIfit program. Then, all subjects’ FA images were processed using TBSS analysis [17,18]: first, each subject’s FA image was nonlinear registrated to the MNI space; second, the mean FA image was calculated and thinned to generate the mean FA skeleton image (FA > 0.2), which represents the centre of white matter tract; third,
PLOS ONE | DOI:10.1371/journal.pone.0119714 March 16, 2015
3 / 12
Covariance Patterns of White Matter in AD
each subject’s transformed FA image was projected to the mean FA skeleton image by calculating the maximum FA values from the nearest tract centre and filling the corresponding position in the skeleton. Lastly, all subjects’ skeletonised FA images were utilised for multivariate ICA.
Multivariate ICA ICA was performed using the Fusion ICA toolbox (FIT) version 2.0c (http://icatb.sourceforge. net/). FIT toolbox is used to examine the shared information between the features and it can also be used to analyze one modality data. In the current study, we did not apply fusion theory and just analyzed DTI data using single modality ICA. The ICA model can be written as X ¼ AS; where X is an observed data matrix, A is an unknown mixing matrix and S is a source matrix that refers to the statistically independent source signals. The ICA model assumes that X is generated by linearly mixing A and S. In this study, the skeletonised FA image of each subject was arrayed into a one-dimensional vector, and the input data matrix X (size: subject number by voxel number) was generated by concatenating the skeletonised FA data (a row vector) of all subjects from the two groups, AD patients and HCs. The initial data matrix X was decomposed using ICA based on the Infomax algorithm to obtain the mixing coefficient matrix A (size: subject number by source number) and the source matrix S (size: source number by voxel number) [21]. Each column of the mixing matrix A (ICA weights) indicates the degree to which each subject expresses the source or network differences between two groups. Each row of the source matrix S represents the source maps exhibiting the inter-subject covariance information.
Statistical analysis A two-sample t-test was performed on each column of the mixing coefficient matrix (ICA weights) to examine the between-group differences (p
