Professional Documents
Culture Documents
Structure and Dynamics of Functional Networks in Child-Onset - Guilherme Ferraz de Arruda and Francisco A. Rodrigues PDF
Structure and Dynamics of Functional Networks in Child-Onset - Guilherme Ferraz de Arruda and Francisco A. Rodrigues PDF
Clinical Neurophysiology
journal homepage: www.elsevier.com/locate/clinph
a r t i c l e i n f o h i g h l i g h t s
Article history:
Cortical networks in healthy and child-onset schizophrenic subjects differ significantly in only a small
Accepted 26 November 2013
Available online xxxx
number of large-scale network properties.
Classification of networks of healthy and schizophrenic subjects yields sensitivity of 90% and specific-
Keywords:
ity of 74%.
Cortical networks Our analysis allows reliable automatic diagnostics in patients with child-onset schizophrenia based
Schizophrenia on cortical network properties and data mining methods.
Child-onset schizophrenia
Complex networks
a b s t r a c t
Objective: Schizophrenia is a neuropsychiatric disorder characterized by cognitive and emotional deficits
and associated with various abnormalities in the organization of neural circuits. It is currently unclear
how and to which extend the global network organization is changed due to such disorder. In this work,
we analyzed cortical networks of healthy subjects and patients with child-onset schizophrenia to address
this issue.
Methods: We performed a comparison of cortical networks extracted from functional MRI data of
patients with schizophrenia and healthy subjects considering their topological and dynamical properties.
Results: Among 54 network measures tested, only four contributed substantially to a discrimination
between the classes of healthy and schizophrenic subjects, with a sensitivity of 90% and specificity of
74%. However, such classes of networks did not differ significantly with respect to the level of network
resilience and synchronization.
Conclusions: Schizophrenic subjects have cortical regions with higher variance of network centrality, but
less modular structure.
Significance: Our findings suggest that it is possible to establish data analysis routines that allow auto-
matic diagnosis of a multifaceted disease like child-onset schizophrenia based on fMRI data of individual
subjects and extracted network properties.
Ó 2013 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights
reserved.
1388-2457/$36.00 Ó 2013 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.clinph.2013.11.036
Please cite this article in press as: de Arruda GF et al. Structure and dynamics of functional networks in child-onset schizophrenia. Clin Neurophysiol
(2014), http://dx.doi.org/10.1016/j.clinph.2013.11.036
2 G.F. de Arruda et al. / Clinical Neurophysiology xxx (2014) xxx–xxx
In addition, analysis of brain network properties derived from characterization and dynamical processes are presented as supple-
fMRI data has been considered for developing methods of non-inva- mentary material. These concepts are considered to develop a net-
sive diagnosis of neurological disorders (Supekar et al., 2008). Diag- work-based approach for diagnosis of child-onset schizophrenia.
nosis of neurological disorders is often hampered by ambiguous
symptoms and the necessity for long-lasting monitoring of the pa- 2.1. Data set
tients. The analysis of brain network properties may provide a way
to facilitate and accelerate diagnosis of such disorders. Neural net- Our study is based on cortical network data obtained from a
work properties derived from resting-state fMRI data have already study by Vértes et al. (Vértes et al., 2012). The authors used rest-
successfully been used for distinguishing between healthy subjects ing-state fMRI to measure low-frequency neural network oscilla-
and patients with different neurological disorders, such as amyo- tions in 140 cortical brain regions at the right hemisphere.
trophic lateral sclerosis (Welsh et al., 2013) major depression (Zeng The data was obtained from a group of adolescent healthy vol-
et al., 2013) antisocial personality disorder (Tang et al., 2013). In our unteers (n = 20, mean age 19.7 years; 11 male) and a group of ado-
study we focused on child-onset schizophrenia, which is a rare form lescent participants with childhood-onset schizophrenia (n = 19,
of schizophrenia having its onset before age of 13. From first med- mean age 18.7 years; 9 male). The subjects were scanned using a
ical consultation to correct diagnosis of child-onset schizophrenia it General Electric Signa MRI scanner operating at 1.5 Tesla. The
can take years (Schaeffer and Ross, 2002) partially because the diag- authors considered only regions in the right hemisphere to facili-
nostic challenges are with differentiating childhood-onset schizo- tate the approximation of the wiring length by the Euclidean dis-
phrenia from other neurological disorders or acute, transient tance between brain regions. Cortical networks of each subject
syndromes (for review see (Masi et al., 2006)). were constructed by determining the functional connectivity from
In case of schizophrenia in general, data obtained from fMRI the correlation between each pair of regional time series using a
suggest that cortical networks of schizophrenic adults are less band-pass filter in order to define the frequency interval in the
small-world type, less hierarchical, less clustered, less efficiently range 0.05 – 0.111 Hz. Binary graphs were constructed by thres-
wired (Bassett et al., 2008; Bullmore and Sporns, 2009) and have holding the wavelet correlation matrix estimated for each subject.
a less hub-dominated configuration (Lynall et al., 2010). Moreover, In order to prevent disconnection of nodes, Vértes et al. considered
Basset et al. (Bassett et al., 2008) showed that multimodal network the minimum spanning tree as a backbone. This procedure resulted
organization of patients with schizophrenia are abnormal, present- in 140 cortical regions whose average time series were used to
ing increased connection distance and reduced hierarchy. In adult construct brain functional networks. These preprocessed data can
schizophrenic patients previous studies already illustrated the use- be downloaded via http://intramural.nimh.nih.gov/chp/articles/
fulness of non-invasive diagnosis based on the analysis of resting- matlab.html. Additional details about the database can be found
state fMRI data. Such analysis of brain network properties achieved in (Vértes et al., 2012).
values for correct classification of up to 94% with 75% accuracy These functional networks were used to develop mathematical
(Tang et al., 2012; Venkataraman et al., 2012; Shen et al., 2010). models of functional human brain networks based on distance
However, similar analysis of fMRI data from patients with child- penalty and connections between regions sharing similar output.
onset schizophrenia is currently lacking. Analysis of data from pa- Vertés and coworkers suggested that some network properties
tients presenting child-onset schizophrenia is very important, can be explained by an economical selection pressure to establish
since it allows early diagnosis and disease monitoring. Moreover, connections between different brain areas. Moreover, the authors
although these studies confirmed various structural differences also verified that clustering and modularity are reduced in schizo-
in cortical networks of schizophrenic subjects, they have not deter- phrenic patients (Vértes et al., 2012).
mined which network properties are most consistently affected by
schizophrenia. The two main objectives of our study are (i) analysis 2.2. Data mining
and evaluation of the topological and dynamical properties of brain
networks of patients with child-onset schizophrenia and healthy Before performing the classification of the cortical networks,
subjects and (ii) determining which network properties are most some pre-processing is required in order to avoid scaling effects,
consistently affected by schizophrenia. since the network measures varies in different ranges. That is,
One of our main findings suggests that the networks in healthy while the degree is an integer larger than zero, the clustering coef-
and schizophrenic subjects differ significantly in only a small num- ficient varies in the interval ½0; 1 (see the supplementary material).
ber of large-scale network properties. On the other hand, healthy One possible method for standardization is the so called z-score
and schizophrenic networks do not differ significantly with respect (Costa and Cesar Jr, 2000; Theodoridis and Koutroumbas, 2003).
to dynamical processes such as resilience to attacks and synchroni- For each observation i; i ¼ 1; . . . ; N, the transformation of the attri-
zation. Therefore, the main structural alterations in the cortical bute xi is performed as
networks of patients with schizophrenia do not affect the general
xi x
network resilience and synchronization of the brain. yi ¼ ; ð1Þ
Finally, we showed that network measures and data-mining
rx
P qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
PN
methods can be useful as an imaging-based diagnosis method to where x ¼ Nj¼1 xj =N and rx ¼ 2
j¼1 ðxj xÞ =N are the average and
distinguish patients presenting child-onset schizophrenia from standard deviation of the values of the measure x, respectively. The
healthy subjects. In this case, a sensitivity of 74% and specificity probability distribution of each transformed measurement, y, pre-
of 90% were obtained. Therefore, our findings suggest that it is pos- sents zero mean and unit variance.
sible to establish a data analysis routine that allows automatic Pre-processing is also necessary for obtaining classes with the
diagnosis of this multifaceted disease based on functional MRI data same number of objects, since a class with the highest number of
of individual subjects and the resulting cortical network properties. elements can be favored in the classification process (Theodoridis
and Koutroumbas, 2003). In this way, we performed a random
sampling, which allows obtaining classes with the same number
2. Materials and methods of elements.
After the pre-processing, we analyzed the topology of functional
In this section, we describe the main concepts adopted in networks via data mining methods in order to find the most impor-
this work regarding data mining. The concepts of network tant features, i.e., those that best discriminate between the two
Please cite this article in press as: de Arruda GF et al. Structure and dynamics of functional networks in child-onset schizophrenia. Clin Neurophysiol
(2014), http://dx.doi.org/10.1016/j.clinph.2013.11.036
G.F. de Arruda et al. / Clinical Neurophysiology xxx (2014) xxx–xxx 3
Please cite this article in press as: de Arruda GF et al. Structure and dynamics of functional networks in child-onset schizophrenia. Clin Neurophysiol
(2014), http://dx.doi.org/10.1016/j.clinph.2013.11.036
4 G.F. de Arruda et al. / Clinical Neurophysiology xxx (2014) xxx–xxx
zero. Consequently these measures did not contribute to the dis- Having identified the most relevant features for discrimination
crimination between the two classes of networks. Note that a mea- of the two classes of networks, we performed a supervised classi-
sure presenting similar values in different data classes has U ¼ 0 fication. We evaluated the data set by considering three different
and v2 ¼ 0 (Theodoridis and Koutroumbas, 2003). Therefore, based supervised classifiers, i.e., naive Bayes, Bayesian network and
on these criteria, out of the 54 original network measures, only C4.5 decision tree (see Section 2.2). We compared the results of
four contributed significantly to a discrimination between the two supervised classifications: (i) a classification based on the four
two classes of networks, which indicates that these properties most relevant measures only and (ii) a classification based on all 54
can be related to abnormalities in cortical networks of patients network measures. As shown in Table 2, the analysis based on the
presenting child-onset schizophrenia. These measures are the var- 4 most important attributes always yielded more precise classifica-
iance of the closeness centrality, the first moment of K-core, the tions. Indeed, the decision tree was the most sensible one to the
maximum modularity and the variance of the accessibility (see selection procedure, improving the accuracy from 44.74% to
Table 1). Our finding that only 4 of the tested measures contributed 71.05%.
to discrimination at the same time indicates that the network
properties quantified by the other 50 measures were similar in net- 3.2. Dynamical processes
works of schizophrenic and healthy subjects. Fig. 1 presents the
probability distributions of these measurements. For better visual- In addition to the structural evaluation, we performed a com-
ization, we considered a kernel density estimation (Gaussian ker- parison between the networks obtained from healthy and schizo-
nel) to smooth the densities (Parzen, 1962; Rosenblatt, 1956). phrenic subjects in terms of two dynamical processes, i.e., the
Cortical networks of schizophrenic subjects tended to present synchronization of coupled oscillators and resilience to failures
higher values of variance of the closeness centrality and accessibil- and attacks.
ity, but small values of average K-core and modularity. Therefore, The synchronization of non-identical oscillators, i.e., the Kuram-
such networks have cortical regions displaying a higher variance oto model, can be considered as being related with lesion recovery
of central properties, but less modular structure. Note that the sep- in cortical networks (Honey and Sporns, 2008). The Kuramoto
arations of the distributions in Fig. 1 agree with the rankings pre- model (Acebrón et al., 2005) was considered by Honey and Sporn
sented in Table 1. Note that U ¼ 0 and v2 ¼ 0 are zero mainly due (Honey and Sporns, 2008) for simulation of the effects of cortical
to the discretization method applied to the dataset during the fea- lesions. This model is described in the supplementary material.
ture ranking procedures. For additional information about the dis- The evolution of the order parameter was similar in healthy and
cretization procedure see Section 2.2 and (Fayyad and Irani, 1993). schizophrenic networks (Fig. 2). The differences observed may be
1200 6
Healthy Healthy
Schizophrenia Schizophrenia
1000 5
800 4
f(E[K−core])
f(V[Cc ])
600 3
400 2
200 1
0 0
0 1 2 3 4 5 2.8 3 3.2 3.4 3.6 3.8 4
−3
V[Cc ] x 10 E[K−core]
(a) (b)
−3
x 10
2
14 Healthy
Healthy
Schizophrenia
12 Schizophrenia
1.5
10
f(V[Acc])
8
f(Q)
1
6
4 0.5
2
0 0
0.4 0.5 0.6 0.7 0.8 0 1000 2000 3000 4000 5000
Q V[Acc]
(c) (d)
Fig. 1. Probability distributions of the network measures with the highest rankings for networks in healthy and schizophrenic subjects: (a) variance of the closeness
centrality, (b) first statistical moment of the K-core, (c) maximum modularity, and (d) variance of the accessibility. We considered a kernel density estimation (Gaussian
kernel) to smooth the densities for better visualization.
Please cite this article in press as: de Arruda GF et al. Structure and dynamics of functional networks in child-onset schizophrenia. Clin Neurophysiol
(2014), http://dx.doi.org/10.1016/j.clinph.2013.11.036
G.F. de Arruda et al. / Clinical Neurophysiology xxx (2014) xxx–xxx 5
Table 2
Percentage of correct classification of networks obtained from healthy and schizophrenic subjects considering 4 or 54 measures. PC is the positive class, H. indicates the healthy
class and S., schizophrenic subjects.
1 1 1
r
0.4 0.4 0.4
Please cite this article in press as: de Arruda GF et al. Structure and dynamics of functional networks in child-onset schizophrenia. Clin Neurophysiol
(2014), http://dx.doi.org/10.1016/j.clinph.2013.11.036
6 G.F. de Arruda et al. / Clinical Neurophysiology xxx (2014) xxx–xxx
1 1
Healthy Healthy
Schizophrenia Schizophrenia
0.8 0.8
0.6 0.6
E/E0
E/E0
0.4 0.4
0.2 0.2
0 0
0 0.2 0.4 0.6 0.8 0 0.1 0.2 0.3 0.4 0.5
Fraction failed Fraction attacked
(a) (b)
Fig. 4. Relative efficiency obtained from (a) random removal of nodes and (b) attacks.
Please cite this article in press as: de Arruda GF et al. Structure and dynamics of functional networks in child-onset schizophrenia. Clin Neurophysiol
(2014), http://dx.doi.org/10.1016/j.clinph.2013.11.036
G.F. de Arruda et al. / Clinical Neurophysiology xxx (2014) xxx–xxx 7
hallucinations or delusions (for review see Masi et al., 2006). He Y, Chen Z, Evans A. Structural insights into aberrant topological patterns of large-
scale cortical networks in Alzheimer’s disease. J Neurosci 2008;28:4756–66.
Therefore, in future studies it has to be tested, whether the analysis
Honey CJ, Sporns O. Dynamical consequences of lesions in cortical networks. Hum
of brain network properties extracted from fMRI data is capable of Brain Mapp 2008;29:802–9.
discriminating between patients that suffer from neurological dis- Jin X, Xu A, Bie R, Guo P. Machine learning techniques and chi-square feature
eases with partially similar phenotypes. As a whole, our findings selection for cancer classification using SAGE gene expression profiles. In:
BioDM. In: Li J, Yang Q, Tan AH, editors. Lecture Notes in Computer Science
indicate that the analysis presented here provides a reliable way 2006;vol. 3916. Springer; 2006. p. 106–15.
to quantify differences in networks of healthy subjects and Liu Y, Liang M, Zhou Y, He Y, Hao Y, Song M, Yu C, Liu H, Liu Z, Jiang T. Disrupted
patients with child-onset schizophrenia, which might a first funda- small-world networks in schizophrenia. Brain 2008;131:945–61.
Lynall M, Bassett D, Kerwin R, McKenna P, Kitzbichler M, Muller U, Bullmore E.
mental step towards an early diagnosis and monitoring of this neu- Functional connectivity and brain networks in schizophrenia. J Neurosci
ropsychiatric disorder. 2010;30:9477–87.
Masi G, Mucci M, Pari C. Children with schizophrenia: clinical picture and
pharmacological treatment. CNS drugs 2006;20:841–66.
Acknowledgments Menon V. Large-scale brain networks and psychopathology: a unifying triple
network model. Trends Cogn Sci 2011;15:483–506.
F.A. Rodrigues would like to acknowledge CNPq (305940/2010- Parzen E. On estimation of a probability density function and mode. Ann Math Stat
1962;33:1065–76.
4) and FAPESP (2010/19440-2) for the financial support given to Rosenblatt M. Remarks on some nonparametric estimates of a density function. Ann
this research. G.F. de Arruda acknowledge Fapesp for sponsorship. Math Stat 1956;27(3(1956)):832–7.
F.A. Rodrigues and L. da F. Costa acknowledge FAPESP (2011/ Schaeffer J, Ross R. Childhood-onset schizophrenia: premorbid and prodromal
diagnostic and treatment histories. J Am Acad Child Adolesc Psychiatry
50761-2) and NAP eScience – PRP – USP.
2002;41:538–45.
The authors declare no competing financial interests. Shannon CE, Weaver W. A mathematical theory of communication. Bell Syst Tech J
1948;27:623.
Shen H, Wang L, Liu Y, Hu D. Discriminative analysis of resting-state functional
Appendix A. Supplementary data connectivity patterns of schizophrenia using low dimensional embedding of
fMRI. Neuroimage 2010;49:3110–21.
Supplementary data associated with this article can be found, in Sporns O. Networks of the Brain. MIT Press; 2010.
Supekar K, Menon V, Rubin D, Musen M, Greicius M. Network analysis of intrinsic
the online version, at http://dx.doi.org/10.1016/j.clinph.2013. functional brain connectivity in Alzheimer’s disease. PLoS Comput Biol
11.036. 2008;4:e1000100.
Tang Y, Jiang W, Liao J, Wang W, Luo A. Identifying individuals with antisocial
personality disorder using resting-state fMRI. PloS one 2013;8:e60652.
References
Tang Y, Wang L, Cao F, Tan L. Identify schizophrenia using resting-state functional
connectivity: an exploratory research and analysis. Biomed Eng Online
Acebrón JA, Bonilla LL, Vicente CJP, Ritort F, Spigler R. The Kuramoto model: a 2012;11:50.
simple paradigm for synchronization phenomena. Rev Mod Phys 2005;77:137. Theodoridis S, Koutroumbas K. Pattern recognition. Academic Press; 2003.
Arenas A, Díaz-Guilera A, Kurths J, Moreno Y, Zhou C. Synchronization in complex Venkataraman A, Whitford T, Westin C, Golland P, Kubicki M. Whole brain resting
networks. Phys Rep 2008;469:93–153. state functional connectivity abnormalities in schizophrenia. Schizophr Res
Barrat A, Barthlemy M, Vespignani A. Dynamical processes on complex 2012;139:7–12.
networks. Cambridge University Press; 2008. Vértes P, Alexander-Bloch A, Gogtay N, Giedd J, Rapoport J, Bullmore E. Simple
Bassett D, Bullmore E. Human brain networks in health and disease. Curr Opin models of human brain functional networks. Proc Natl Acad Sci USA
Neurol 2009;22:340. 2012;109:5868–73.
Bassett D, Bullmore E, Verchinski B, Mattay V, Weinberger D, Meyer-Lindenberg A. Welsh R, Jelsone-Swain L, Foerster B. The utility of independent component analysis
Hierarchical organization of human cortical networks in health and and machine learning in the identification of the amyotrophic lateral sclerosis
schizophrenia. J Neurosci 2008;28:9239–48. diseased brain. Front Hum Neurosci 2013;7(251):1–9. http://dx.doi.org/
Bullmore E, Sporns O. Complex brain networks: graph theoretical analysis of 10.3389/fnhum.2013.00251.
structural and functional systems. Nat Rev Neurosci 2009;10:186–98. Witten IH, Frank E. Data mining: practical machine learning tools and
Chen Z, He Y, Rosa-Neto P, Germann J, Evans A. Revealing modular architecture of techniques. Morgan Kaufmann; 2005.
human brain structural networks by using cortical thickness from MRI. Cerebral Yang Y, Pedersen JO. A comparative study on feature selection in text
Cortex 2008;18:2374–81. categorization. In: Fisher DH, editor. Proceedings of 14th International
Costa LF, Cesar Jr R. Shape analysis and classification: theory and practice. CRC Press conference on machine learning, ICML-97. San Francisco, US: Morgan
Inc.; 2000. Kaufmann Publishers; 1997. p. 412–20.
Demetrius L, Manke T. Robustness and network evolution—an entropic principle. Yu L, Liu H. Feature selection for high-dimensional data: a fast correlation-based
Phys A 2005;346:682–96. filter solution. In: Proceedings of the Twentieth International Conference on
Fayyad UM, Irani KB. Multi-interval discretization of continuous-valued attributes Machine Learning. AAAI Press; 2003. p. 856–63.
for classification learning. In: Bajcsy R, editor. IJCAI. Morgan Kaufmann; 1993. p. Zeng, L., Shen, H., Liu, L., Hu, D., 2013. Unsupervised classification of major
1022–9. depression using functional connectivity. MRI. Hum Brain Mapp. http://
Hall M, Frank E, Holmes G, Pfahringer B, Reutemann P, Witten IH. The WEKA data dx.doi.org/10.1002/hbm.22278.
mining software: an update. SIGKDD Explorations 2009;11(1):10–8.
Please cite this article in press as: de Arruda GF et al. Structure and dynamics of functional networks in child-onset schizophrenia. Clin Neurophysiol
(2014), http://dx.doi.org/10.1016/j.clinph.2013.11.036