Beruflich Dokumente
Kultur Dokumente
6, SEPTEMBER 2017
Abstract — Brain network plays an important role in repre- but measurable memory loss, is considered as the transition
senting abnormalities in Alzheimers disease (AD) and mild stage between health controls (HC) and dementia. Patients
cognitive impairment (MCI), which includes MCIc (MCI con- with MCI have been reported more prone to progress to AD
verted to AD) and MCInc (MCI not converted to AD). In our
previous study, we proposed an AD classification approach than HC [1], [2]. The symptoms of dementia seriously affect
based on individual hierarchical networks constructed with patient’s daily life. As the major form of dementia, AD affects
3D texture features of brain images. However, we only used 26.6 million of people worldwide in 2006 and it is predicted to
edge features of the networks without node features of affect 1 in 85 people by 2050 [3]. Therefore, precise prediction
the networks. In this paper, we propose a framework of and diagnosis of AD, especially at its early warning stage such
the combination of multiple kernels to combine edge fea-
tures and node features for AD classification. An evaluation as MCI, have become a crucial step to retard or even avoid
of the proposed approach has been conducted with MRI dementia.
images of 710 subjects (230 health controls (HC), 280 MCI With the rapid development of brain neuroimaging tech-
(including 120 MCIc and 160 MCInc), and 200 AD) from niques, such as magnetic resonance imaging (MRI), functional
the Alzheimer’s disease neuroimaging initiative database MRI (fMRI), diffusion tensor imaging (DTI), etc., brain struc-
by using ten-fold cross validation. Experimental results
show that the proposed method is not only superior to tural and functional changes can be examined with various
the existing AD classification methods, but also efficient acquired images [4]–[7]. Owning to its easy access in clinical
and promising for clinical applications for the diagnosis of settings, MRI receives the most attention compared with other
AD via MRI images. Furthermore, the results also indicate imaging techniques. Studies using MRI have showed that
that 3D texture could detect the subtle texture differences AD patients are closely related to brain atrophy [8]–[13].
between tissues in AD, MCI, and HC, and texture features of
MRI images might be related to the severity of AD cognitive For example, hippocampal atrophy is considered to be the sig-
impairment. These results suggest that 3D texture is a useful nature of AD and it is therefore used as a clinical marker [4];
aid in AD diagnosis. medial temporal lobe (MTL) atrophy as assessed by using MRI
Index Terms — Alzheimer’s disease, 3D texture, individual has proven to be an effective clinical aid in the early diagnosis
hierarchical network, multiple kernel learning, classifica- of AD [8], [9]; the location and degree of gray matter atrophy
tion. can be accurately visualized with MRI, which are used for the
clinical diagnosis of AD [12].
I. I NTRODUCTION Texture features play an important role in image analysis.
1536-1241 © 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
LIU et al.: AD CLASSIFICATION BASED ON INDIVIDUAL HIERARCHICAL NETWORKS CONSTRUCTED WITH 3-D TEXTURE FEATURES 429
TABLE II
C EREBELLAR ROIs B ASED ON AAL ATLAS IN THE P RESENT S TUDY
• Several ROIs in S1 are combined into a ROI if the 2) Construction of the Individual Hierarchical Network: A net-
first two digits and the last digit of their No. are the work is typically defined as G = (V, E), where V is the set
same. For example, Cingulum_Ant_L (4001), Cingu- of nodes and E is the set of edges. In this study, we assume
lum_Mid_L (4011) and Cingulum_Post_L (4021) are that a node is a cerebral region as defined by the above-
considered as a ROI. Based on this rule, the atlas includes mentioned four atlas (S1, S2, S3 and S4) and an edge is a link
54 ROIs and is denoted as S2. between two ROIs from the above-mentioned four atlas. Based
• Similar to the rule of the generation of S2. Some on this assumption, we can construct an individual hierarchical
ROIs in S1 are combined into a ROI if the first and network for each subject as shown in Fig.1 [24].
the last digit of their No. are the same. For example, For each node (ROI) of the individual hierarchical net-
Cingulum_Ant_L (4001), Cingulum_Mid_L (4011), work, we extract six 3D texture properties to represent it,
Cingulum_Post_L (4021), Hippocampus_L (4101), including energy (ENE), contrast (CON), inverse difference
ParaHippocampal_L (4111) and Amygdala_L (4201) moment (IDM), entropy (ENT), difference variance (DVA) and
are considered as a ROI. Based on this rule, the atlas difference entropy (DEN). For each node of each 3D texture
includes 14 ROIs and is denoted as S3. property, we firstly calculate the 3D texture property of the
• The whole brain is regarded as a ROI. The atlas only ROI in 0°, 45°, 90° and 135° directions [40] by using
includes 1 ROI and is denoted as S4. gray level co-occurrence matrices (GLCM) method and then
LIU et al.: AD CLASSIFICATION BASED ON INDIVIDUAL HIERARCHICAL NETWORKS CONSTRUCTED WITH 3-D TEXTURE FEATURES 431
D. Feature Selection
In general, some of the features are noisy or redundant,
and even negatively affect classification. Therefore, feature
selection is an essential step before performing classification.
The feature ranking method [41]–[44] has been widely applied
in feature selection. F-score [45] is a simple and quite
effective method which evaluates the discrimination between
two different datasets. In this study, the F-score method is
used to calculate the discrimination of each feature. The
F-score of the i -th feature is calculated as following:
(+) 2 (−) 2
(x i − x i ) + (x i − xi )
F(i ) =
Fig. 1. The individual hierarchical network.
n+
(+) (+) 2
n−
(−) (−) 2
1
n + −1 (x k,i − x i ) + 1
n − −1 (x k,i − x i )
k=1 k=1
compute the average texture property of the ROI in four (8)
directions. The six 3D texture properties are calculated as where n + and n − are denoted as the number of positive and
following: (+) (−)
negative samples respectively; x i , x i and x i are the aver-
ENE = p(i, j )2 (1) age of the i -th feature of the all, positive and negative samples
(+)
i=1 j =1 respectively; x k,i is the i -th feature of the k-th positive sample,
(−)
and x k,i is the i -th feature of the k-th negative sample. The
CON = (i − j )2 p(i, j ) (2) numerator indicates the discrimination between the positive
i=1 j =1 and negative samples, and the denominator indicates the one
1 within each of the two samples. The larger the F-score
I DM = p(i, j ) (3) is, the more likely this feature is discriminative. Therefore,
i=1 j =1
1 + (i − j )2 by using this score, the features with good discriminative
power are selected to AD classification.
E NT = − p(i, j ) log( p(i, j )) (4)
i=1 j =1
E. Classification and Evaluation
2
DV A = (l − u x−y ) px−y (l) (5) Multiple Kernel Learning (MKL) is a promising family
l=0 of machine learning algorithms using multiple kernel func-
tions for various challenging data mining tasks [46]–[49].
DE N = − px−y (l) log( p x−y (l)) (6) In this study, we firstly use MKBoost algorithm proposed by
l=0
Xia and Hoi [49] to obtain the best classification accuracy and
p(i, j ) is the value of GLCM, px−y (l) =
where the corresponding classifier for each feature set, such as ENE
p(i, j ), u x−y is the average value of px−y (l), where features, CON features, IDM features, ENT features, DVA
i=1 j =1
l = |i − j|. For more details with GLCM, please see [40]. features, DEN features and edge features. For each feature
Based on the above analysis, a 6-dimensional vector is set, we create a set of 13 base kernels, i.e.,
• Gaussian kernels with 10 different
generated to represent each node of the individual hierarchical
network. In this study, to obtain the weight of each edge widths ( 2−4 , 2−3 , . . . , 24 , 25 ) on all features.
of the individual hierarchical network, we calculate Pearson • Polynomial kernels of degree 1 to 3 on all features.
correlation coefficients (PCC) between two nodes below: In our study, we set the total number of boosting trials T
E(ab) − E(a)E(b) to 100, the boosting sampling ratio to 0.2, and the sam-
PCC(a, b) = (7) pling decay factor to 2−5 for MKBoost-S2. Support vector
E(a2 ) − E 2 (a) E(b2 ) − E 2 (b) machine (SVM) is adopted to perform classification and
where a, b are node vectors, E(·) is the average value of a implemented by the LIBSVM library [50], with a default value
vector. for the parameter C (= 1).
At this point, we can obtain node features and edge fea- Then, we use a simple framework of the combination
tures for each individual hierarchical network. Node features of multiple kernels (CK) to generate a final classifier as
include ENE features, CON features, IDM features, ENT shown in Fig.2. Suppose there are N kernel functions,
features, DVA features and DEN features, which make up a {k1 (·, ·), . . . , k N (·, ·)}, corresponding to N feature sets, X =
159(90 + 54 + 14 + 1) dimensional vector for each subject. {X (1) , . . . , X (N) }, where matrix X (n) represents samples using
432 IEEE TRANSACTIONS ON NANOBIOSCIENCE, VOL. 16, NO. 6, SEPTEMBER 2017
Fig. 3. A flowchart for novel individual hierarchical network-based feature representation for Alzheimer’s disease classification.
Fig. 4. Classification accuracy for four types. The features are ranked according to F -score in descending order.
Table IV) and the corresponding classifier. This work is the Based on the above analysis, we can demonstrate that our
same as [24]. proposed feature representation based on individual hierar-
We assume that the combination of node features and chical networks is efficient for improving AD classification
edge features may improve the classification accuracy of AD. performance.
Therefore, the best classification accuracies of node fea-
tures and edge features are used as the weights to com- C. Comparison With Existing Classification Methods
bine their classifiers (namely N-CK and E-MK) as the final In order to validate the superiority of our proposed method,
classifier (NECK) for the above four types of AD clas- we compare the six existing methods for AD classification as
sification. The results of NECK are shown in Table IV. follows:
As shown in Table IV, we can easily find that the classification • Chupin et al. [51] proposed a segmentation method to
performance (AUC) of NECK is superior to both N-CK segment hippocampus and used the obtained hippocampal
and E-MK. volumes for AD classification.
434 IEEE TRANSACTIONS ON NANOBIOSCIENCE, VOL. 16, NO. 6, SEPTEMBER 2017
TABLE V
C OMPARISON TO E XISTING W ORKS FOR T1
TABLE VI
C OMPARISON TO E XISTING W ORKS FOR T2
TABLE VII
C OMPARISON TO E XISTING W ORKS FOR T3
Fig. 5. Classification accuracy for four types. The features are ranked
according to F -score in descending order.
• Ahmed et al. [52] proposed to use hippocampus visual • Liu et al. [24] proposed to use edge features of the
features and cerebrospinal fluid (CSF) volume for AD hierarchial networks for AD classification.
classification. For fair comparisons, the experimental images of these meth-
• Dai et al. [23] proposed to construct an individual net- ods are the same as those of our proposed method. Although
work based on mean cortical thickness and used the Suk et al. [53] used two modal images: MRI and positron
network edge features for AD classification. emission tomography (PET), only the results based on MRI
• Suk et al. [53] proposed to use DBM (Deep Boltzmann images are reported in this study. The comparative results for
Machine) to find a latent hierarchical feature representa- T1, T2, T3 and T4 are shown in Tables V, VI, VII and VIII
tion from a 3D patch for AD classification. respectively.
• Khedher et al. [54] proposed to use GM and WM for As can be seen from Tables V to VIII, our proposed method
AD classification. substantially outperforms the other six existing methods in
LIU et al.: AD CLASSIFICATION BASED ON INDIVIDUAL HIERARCHICAL NETWORKS CONSTRUCTED WITH 3-D TEXTURE FEATURES 435
[24] J. Liu, M. Li, W. Lan, F.-X. Wu, Y. Pan, and J. Wang, “Classi- [48] Z. Xu, R. Jin, I. King, and M. Lyu, “An extended level method for
fication of Alzheimer’s disease using whole brain hierarchical net- efficient multiple kernel learning,” in Proc. Adv. Neural Inf. Process.
work,” IEEE/ACM Trans. Comput. Biol. Bioinf., to be published, doi: Syst., 2009, pp. 1825–1832.
10.1109/TCBB.2016.2635144. [49] H. Xia and S. C. H. Hoi, “MKBoost: A framework of multiple kernel
[25] N. Tzourio-Mazoyer et al., “Automated anatomical labeling of activa- boosting,” IEEE Trans. Knowl. Data Eng., vol. 25, no. 7, pp. 1574–1586,
tions in SPM using a macroscopic anatomical parcellation of the MNI Jul. 2013.
MRI single-subject brain,” NeuroImage, vol. 15, no. 1, pp. 273–289, [50] C.-C. Chang and C.-J. Lin, “LIBSVM: A library for support vec-
2002. tor machines,” ACM Trans. Intell. Syst. Technol., vol. 2, no. 3,
[26] M. C. Carrillo, L. J. Bain, G. B. Frisoni, and M. W. Weiner, “Worldwide pp. 27:1–27:27, 2011. [Online]. Available: http://www.csie.ntu.edu.
Alzheimer’s disease neuroimaging initiative,” Alzheimer’s Dementia, tw/~cjlin/libsvm
vol. 8, no. 4, pp. 337–342, Jul. 2012. [51] M. Chupin et al., “Fully automatic hippocampus segmentation and
[27] J. G. Sled, A. P. Zijdenbos, and A. C. Evans, “A nonparamet- classification in Alzheimer’s disease and mild cognitive impairment
ric method for automatic correction of intensity nonuniformity in applied on data from ADNI,” Hippocampus, vol. 19, no. 6, pp. 579–587,
MRI data,” IEEE Trans. Med. Imag., vol. 17, no. 1, pp. 87–97, 2009.
Feb. 1998. [52] O. B. Ahmed et al., “Classification of Alzheimer’s disease subjects
[28] S. M. Smith, “Fast robust automated brain extraction,” Human Brain from MRI using hippocampal visual features,” Multimedia Tools Appl.,
Mapping, vol. 17, no. 3, pp. 143–155, 2002. vol. 74, no. 4, pp. 1249–1266, Feb. 2015.
[29] M. Jenkinson and S. Smith, “A global optimisation method for robust [53] H.-I. Suk, S.-W. Lee, and D. Shen, “Hierarchical feature representation
affine registration of brain images,” Med. Image Anal., vol. 5, no. 2, and multimodal fusion with deep learning for AD/MCI diagnosis,”
pp. 143–156, Jun. 2001. NeuroImage, vol. 101, pp. 569–582, Nov. 2014.
[30] M. Jenkinson, P. Bannister, M. Brady, and S. Smith, “Improved opti- [54] L. Khedher, J. Ramírez, J. M. Górriz, A. Brahim, and F. Segovia, “Early
mization for the robust and accurate linear registration and motion diagnosis of Alzheimer?s disease based on partial least squares, principal
correction of brain images,” NeuroImage, vol. 17, no. 2, pp. 825–841, component analysis and support vector machine using segmented MRI
Oct. 2002. images,” Neurocomputing, vol. 151, pp. 139–150, Mar. 2015.
[31] B. Magnin et al., “Support vector machine-based classification of
Alzheimer’s disease from whole-brain anatomical MRI,” Neuroradiol-
ogy, vol. 51, no. 2, pp. 73–83, Feb. 2009.
[32] D. Zhang, Y. Wang, L. Zhou, H. Yuan, and D. Shen, “Multimodal
classification of Alzheimer’s disease and mild cognitive impairment,”
NeuroImage, vol. 55, no. 3, pp. 856–867, Apr. 2011.
[33] X. Peng, J. Wang, W. Peng, F.-X. Wu, and Y. Pan, “Protein- Jin Liu is currently pursuing the Ph.D. degree
protein interactions: Detection, reliability assessment and applica- with the School of Information Science and Engi-
tions,” Briefings Bioinform., p. bbw066, 2016. [Online]. Available: neering, Central South University, Changsha,
https://doi.org/10.1093/bib/bbw066 China. His current research interests include
[34] H. Luo et al., “Drug repositioning based on comprehensive similarity medical image analysis, bioinformatics, and
measures and bi-random walk algorithm,” Bioinformatics, vol. 32, machine learning.
no. 17, pp. 2664–2671, 2016.
[35] M. Li, P. Ni, X. Chen, J. Wang, F. Wu, and Y. Pan, “Construction
of refined protein interaction network for predicting essential pro-
teins,” IEEE/ACM Trans. Comput. Biol. Bioinf., to be published, doi:
10.1109/TCBB.2017.2665482.
[36] Y. Liu et al., “Disrupted small-world networks in schizophrenia,” Brain,
vol. 131, no. 4, pp. 945–961, 2008.
[37] R. Vandenberghe et al., “18 F-flutemetamol amyloid imaging in
Alzheimer disease and mild cognitive impairment: A phase 2 trial,” Ann.
Neurol., vol. 68, no. 3, pp. 319–329, Sep. 2010.
Jianxin Wang (M’01–SM’14) received the
[38] J. Wang et al., “Disrupted functional brain connectome in individuals
B.Eng. and M.Eng. degrees in computer engi-
at risk for Alzheimer’s disease,” Biol. Psychiatry, vol. 73, no. 5,
neering and the Ph.D. degree in computer sci-
pp. 472–481, Mar. 2013.
ence from Central South University, Changsha,
[39] J. Zhang, L. Zhou, L. Wang, and W. Li, “Functional brain network
China, in 1992, 1996, and 2001, respectively. He
classification with compact representation of SICE matrices,” IEEE
is currently the Chair and a Professor with the
Trans. Biomed. Eng., vol. 62, no. 6, pp. 1623–1634, Jun. 2015.
Department of Computer Science, Central South
[40] R. M. Haralick, K. Shanmugam, and I. H. Dinstein, “Textural fea- University. His current research interests include
tures for image classification,” IEEE Trans. Syst., Man, Cybern., Syst., algorithm analysis and optimization, parameter-
vol. SMC-3, no. 6, pp. 610–621, 1973. ized algorithm, bioinformatics, and computer net-
[41] C. Chu, A.-L. Hsu, K.-H. Chou, P. Bandettini, and C. Lin, “Does feature work.
selection improve classification accuracy? Impact of sample size and
feature selection on classification using anatomical magnetic resonance
images,” NeuroImage, vol. 60, no. 1, pp. 59–70, Mar. 2012.
[42] F. Liu, H.-I. Suk, C.-Y. Wee, H. Chen, and D. Shen, “High-order graph
matching based feature selection for Alzheimer’s disease identifica-
tion,” in Medical Image Computing and Computer-Assisted Intervention.
Springer, 2013, pp. 311–318. Bin Hu received the M.Sc. degree in computer
[43] J. Gao et al., “A novel approach for lie detection based on f-score and science from the Beijing University of Technology
extreme learning machine,” PLoS ONE, vol. 8, no. 6, p. e64704, 2013. and the Ph.D. degree in computer science from
[44] F. Liu et al., “Multivariate classification of social anxiety disorder using the Institute of Computing Technology, Chinese
whole brain functional connectivity,” Brain Struct. Funct., vol. 220, Academy of Science. He is currently a Professor
no. 1, pp. 101–115, Jan. 2015. and the Dean of School of Information Science
[45] Y.-W. Chen and C.-J. Lin, “Combining svms with various feature and Engineering, Lanzhou University. He is the
selection strategies,” in Feature Extraction. Berlin, Germany: Springer, Leader of the Intelligent Contextual Computing
2006, pp. 315–324. Group, Pervasive Computing Center, the Reader
[46] S. Sonnenburg, G. Rätsch, C. Schäfer, and B. Schölkopf, “Large scale with Birmingham City University, and a Visiting
multiple kernel learning,” J. Mach. Learn. Res., vol. 7, pp. 1531–1565, Professor with the Beijing University of Posts and
Jul. 2006. Telecommunications, China, and ETH Zürich, Switzerland. His research
[47] A. Rakotomamonjy, F. Bach, S. Canu, and Y. Grandvalet, “SimpleMKL,” interests include pervasive computing, psycho-physiological computing,
J. Mach. Learn. Res., vol. 9, pp. 2491–2521, Nov. 2008. cooperative work, and semantic Web.
LIU et al.: AD CLASSIFICATION BASED ON INDIVIDUAL HIERARCHICAL NETWORKS CONSTRUCTED WITH 3-D TEXTURE FEATURES 437
Fang-Xiang Wu (M’06–SM’11) received the Yi Pan received the B.Eng. and M.Eng. degrees
B.Sc. degree and the M.Sc. degree in applied in computer engineering from Tsinghua Univer-
mathematics, from the Dalian University of Tech- sity, China, in 1982 and 1984, respectively, and
nology, Dalian, China, in 1990 and 1993, respec- the Ph.D. degree in computer science from the
tively, the first Ph.D. degree in control theory and University of Pittsburgh, USA, in 1991. His profile
its applications from Northwestern Polytechnical has been featured as a distinguished alumnus in
University, Xi’an, China, in 1998, and the second both Tsinghua Alumni Newsletter and University
Ph.D. degree in biomedical engineering from the of Pittsburgh CS Alumni Newsletter. He joined
University of Saskatchewan (U of S), Saskatoon, Georgia State University, USA, in 2000 and was
Canada, in 2004. From 2004 to 2005, he was promoted to a Full Professor in 2004, named a
a Post-Doctoral Fellow with the Laval University Distinguished University Professor in 2013 and
Medical Research Center, Sainte-Foy, QC, Canada. He is currently a Pro- designated a Regents’ Professor (the highest recognition given to a
fessor with the Division of Biomedical Engineering and the Department Faculty Member by the University System of Georgia) in 2015. He served
of Mechanical Engineering, U of S. He has authored over 260 technical as the Chair of the Computer Science Department from 2005 to 2013.
papers in refereed journals and conference proceedings. His current He is currently a Regents’ Professor of Computer Science and an Interim
research interests include computational and systems biology, genomic Associate Dean and the Chair of Biology with Georgia State University.
and proteomic data analysis, biological system identification and para- He is also a Visiting Changjiang Chair Professor with Central South Uni-
meter estimation, and the applications of control theory to biological versity, China. He has authored over 330 papers, including over 180 SCI
systems. He is serving as the Editorial Board Member of three inter- journal papers and 60 IEEE/ACM Transactions papers. In addition, he
national journals, the Guest Editor of several international journals, has edited/authored 40 books. His research interests include parallel and
and as the Program Committee Chair or a member of several interna- cloud computing, wireless networks, and bioinformatics. He is a recipient
tional conferences. He has also reviewed papers for many international of many awards, including the IEEE Transactions Best Paper Award,
journals. four other international conference or journal Best Paper Awards, four
IBM Faculty Awards, two JSPS Sen-ior Invitation Fellowships, the IEEE
BIBE Outstanding Achievement Award, the NSF Research Opportunity
Award, and the AFOSR Summer Faculty Research Fellowship. His
work has been cited over 6500 times. He has served as an Editor-in-
Chief or the Editorial Board Member for 15 journals, including seven
IEEE Transactions. He has organized many international conferences
and delivered keynote speeches at over 50 international conferences
around the world.