**5. Concluding remarks**

In this article, we reviewed the application of graphical models for inferring brain connectivity from fMRI data. We have described and provided signal processing solutions for the challenges raised in this highly interdisciplinary and innovating research field related to model reliability, generality and interpretability.

The importance of error control during brain network structure learning has been increasingly recognized, with a series of papers being published since 2005. These papers proposed solutions from the perspective of both classical and Bayesian statistics, and provided some theoretical conclusions. Because brain regions are not just algebraically isolated variables, but rather located in a three-dimension space with complex geometric structure, a desirable future direction is to exploit this geometric information for improving the error control.

Group analysis is a frequently encountered requirement in biomedical research. Graphical models introduce inter-subject diversity at both the parameter level and the structure level. Most existing methods can be considered to take a two-level framework: a lower level of models for each individual subject, and a group level integrating individual models and describing inter-subject commonality and diversity. Being modular, incrementally updatable, and scalable is highly desirable, yet not well implemented features for current group analysis. Network analysis is an important post-processing for extracting interpretable and human-understandable information from graphical models. Network concepts such as centrality, modularity, connection patterns, the "small-world", "scale-free" property have been actively explored in the analysis of brain connectivity. As various calculation methods could be proposed to quantify the same network concept, it is of great theoretic and practical importance to further pursue criteria for rigorously comparing the discriminability of network measures.

### **6. Appendix**

## **Software and databases**

The interest on modeling brain connectivity using fMRI has been experiencing an increasing, important growth in the signal processing community during the last decade. One of the factors of this success is the availability of public-available software and databases. As a reference for interested readers, here we provide an overview of several widely used computer programs related to fMRI brain connectivity analysis. This list is by no means complete.

• **Statistical Parametric Mapping (SPM)**: Developed by the Wellcome Trust Centre for Neuroimaging.

Website: http://www.fil.ion.ucl.ac.uk/spm

Brief description: The SPM software package, probably the most popular one, has been designed for the analysis of brain imaging data sequences. The sequences can be a series of images from different cohorts, or time-series from the same subject. The current release is designed for the analysis of fMRI, PET, SPECT, EEG and MEG.

• **LONI Software**: Developed by the Laboratory of Neuro Imaging at the University of California, Los Angeles.

Website: http://www.loni.ucla.edu/Software

Brief description: The popular LONI Software is a comprehensive library for neuroimaging analysis, including pipelines for automated processing, web-based applications, tools for image processing and visualization, etc.


Website: http://www.brain-connectivity-toolbox.net

Brief description: This toolbox provides an access to a large selection of complex network measures in Matlab. Such measures aim to characterize brain connectivity by neuro-biologically meaningful statistics, and are used in the description of structural and functional connectivity datasets.

There are normally fMRI datasets associated with the above software. Here we also briefly mention a few publicly-available fMRI databases. Details related with the experiment, design and data content are available in the associated website links.

• **fMRI Data Center (fMRIDC)**: Funded by the National Science Foundation, the W. M. Keck Foundation.

Website: http://www.fmridc.org


#### **7. References**

18 Will-be-set-by-IN-TECH

In this article, we reviewed the application of graphical models for inferring brain connectivity from fMRI data. We have described and provided signal processing solutions for the challenges raised in this highly interdisciplinary and innovating research field related to

The importance of error control during brain network structure learning has been increasingly recognized, with a series of papers being published since 2005. These papers proposed solutions from the perspective of both classical and Bayesian statistics, and provided some theoretical conclusions. Because brain regions are not just algebraically isolated variables, but rather located in a three-dimension space with complex geometric structure, a desirable future

Group analysis is a frequently encountered requirement in biomedical research. Graphical models introduce inter-subject diversity at both the parameter level and the structure level. Most existing methods can be considered to take a two-level framework: a lower level of models for each individual subject, and a group level integrating individual models and describing inter-subject commonality and diversity. Being modular, incrementally updatable, and scalable is highly desirable, yet not well implemented features for current group analysis. Network analysis is an important post-processing for extracting interpretable and human-understandable information from graphical models. Network concepts such as centrality, modularity, connection patterns, the "small-world", "scale-free" property have been actively explored in the analysis of brain connectivity. As various calculation methods could be proposed to quantify the same network concept, it is of great theoretic and practical importance to further pursue criteria for rigorously comparing the discriminability of network

The interest on modeling brain connectivity using fMRI has been experiencing an increasing, important growth in the signal processing community during the last decade. One of the factors of this success is the availability of public-available software and databases. As a reference for interested readers, here we provide an overview of several widely used computer programs related to fMRI brain connectivity analysis. This list is by no means complete.

• **Statistical Parametric Mapping (SPM)**: Developed by the Wellcome Trust Centre for

Brief description: The SPM software package, probably the most popular one, has been designed for the analysis of brain imaging data sequences. The sequences can be a series of images from different cohorts, or time-series from the same subject. The current release

• **LONI Software**: Developed by the Laboratory of Neuro Imaging at the University of

Brief description: The popular LONI Software is a comprehensive library for neuroimaging analysis, including pipelines for automated processing, web-based

Website: http://www.fil.ion.ucl.ac.uk/spm

Website: http://www.loni.ucla.edu/Software

is designed for the analysis of fMRI, PET, SPECT, EEG and MEG.

applications, tools for image processing and visualization, etc.

direction is to exploit this geometric information for improving the error control.

**5. Concluding remarks**

measures.

**6. Appendix**

**Software and databases**

Neuroimaging.

California, Los Angeles.

model reliability, generality and interpretability.


Koivisto, M. & Sood, K. (2004). Exact Bayesian structure discovery in Bayesian networks, *J*

Graphical Models of Functional MRI Data for Assessing Brain Connectivity 395

Kötter, R., Reid, A. T., Krumnack, A., Wanke, E. & Sporns, O. (2007). Shapley ratings in brain

Lauritzen, S. L. (1996). *Graphical Models*, Clarendon Press, Oxford University Press, Oxford,

Li, J. & Wang, Z. J. (2009). Controlling the false discovery rate of the association/causality structure learned with the PC algorithm, *J MACH LEARN RES* pp. 475–514. Li, J., Wang, Z. J., Palmer, S. J. & McKeown, M. J. (2008). Dynamic bayesian network modeling of fMRI: A comparison of group-analysis methods, *NeuroImage* 41(2): 398 – 407. Listgarten, J. & Heckerman, D. (2007). Determining the number of non-spurious arcs in

*Proceedings of the 23rd Conference on Uncertainty in Artificial Intelligence*. Madigan, D., York, J. & Allard, D. (1995). Bayesian graphical models for discrete data, *INT*

differences and commonalities, *NeuroImage* 17(3): 1459–1469.

resonance images., *Neuroimage* 20(2): 1346–1355.

McKeown, M. J. (2000). Cortical activation related to arm-movement combinations, *Muscle &*

Mechelli, A., Penny, W. D., Price, C. J., Gitelman, D. R. & Friston, K. J. (2002). Effective

Milo, R., Shen-Orr, S., Itzkovitz, S., Kashtan, N., Chklovskii, D. & Alon, U. (2002). Network motifs: Simple building blocks of complex networks, *Science* 298(5594): 824–827. Mumford, J. A. & Nichols, T. (2006). Modeling and inference of multisubject fMRI data, *IEEE*

Neumann, J. & Lohmann, G. (2003). Bayesian second-level analysis of functional magnetic

Niculescu-Mizil, A. & Caruana, R. (2007). Inductive transfer for Bayesian network structure

Palmer, S. J., Eigenraam, L., Hoque, T., McCaig, R. G., Troiano, A. & McKeown, M. J. (2009).

Price, C. J. & Friston, K. J. (2002). Degeneracy and cognitive anatomy, *Trends in Cognitive*

Richardson, T. (2003). Markov properties for acyclic directed mixed graphs, *Scandinavian*

Roweis, S. & Ghahramani, Z. (1999). A unifying review of linear Gaussian models, *Neural*

Schlösser, R. G. M., Wagner, G., Koch, K., Dahnke, R., Reichenbach, J. R. & Sauer, H. (2008).

Schwarz, G. (1978). Estimating the dimension of a model, *The Annals of Statistics* 6(2): 461–464.

Song, C., Havlin, S. & Makse, H. A. (2005). Self-similarity of complex networks, *Nature*

Spirtes, P., Glymour, C. & Scheines, R. (2001). *Causation, Prediction, and Search*, The MIT Press.

Fronto-cingulate effective connectivity in major depression: a study with fMRI and

movements in Parkinson's disease., *Neuroscience* 158(2): 693–704.

dynamic causal modeling., *Neuroimage* 43(3): 645–655.

learning, *Proceedings of the 11th International Conference on AI and Statistics (AISTATS*

Levodopa-sensitive, dynamic changes in effective connectivity during simultaneous

a learned DAG model: Investigation of a Bayesian and a frequentist approach,

connectivity and intersubject variability: Using a multisubject network to test

*MACH LEARN RES* 5: 549–573.

*STAT REV* 63(2): 215–232.

*Eng Med Biol* 25(2): 42–51.

*Sciences* 6(10): 416–421.

*Journal of Statistics* 30(1): 145–157.

*Computation* 11(2): 305–345.

Short Communications.

433(7024): 392–395.

*'07)*.

*Nerve* 23(S9): S19–S25.

New York.

networks., *Front Neuroinformatics* 1: 2.


20 Will-be-set-by-IN-TECH

Benjamini, Y. & Yekutieli, D. (2001). The control of the false discovery rate in multiple testing

Bonacich, P. (1972). Factoring and weighting approaches to status scores and clique

Bullmore, E., Fadili, J., Maxim, V., Sendur, L., Whitcher, B., Suckling, J., Brammer, M. &

Bullmore, E. & Sporns, O. (2009). Complex brain networks: graph theoretical analysis of

Calhoun, V., Adali, T., Pearlson, G. & Pekar, J. (2001). A method for making group inferences

Cao, J. & Worsley, K. (1999). The geometry of correlation fields with an application to

Catani, M. & ffytche, D. H. (2005). The rises and falls of disconnection syndromes., *Brain*

Corneil, D. & Gotlieb, C. (1970). An efficient algorithm for graph isomorphism, *Journal of the*

Edelman, G. M. & Gally, J. A. (2001). Degeneracy and complexity in biological systems, *PNAS*

Estrada, E. & Rodrguez-Velzquez, J. A. (2005). Subgraph centrality in complex networks, *Phys.*

Freeman, L. C. (1977). A set of measures of centrality based on betweenness, *Sociometry*

Friedman, N. & Koller, D. (2003). Being Bayesian about network structure. A Bayesian

Friston, K. J., Harrison, L. & Penny, W. (2003). Dynamic causal modelling, *NeuroImage*

Girvan, M. & Newman, M. E. J. (2002). Community structure in social and biological

Granger, C. W. J. (Aug., 1969). Investigating causal relations by econometric models and

Harville, D. A. (1977). Maximum likelihood approaches to variance component estimation

Heckerman, D., Geiger, D. & Chickering, D. M. (1995). Learning Bayesian networks: The combination of knowledge and statistical data, *Machine Learning* 20(3): 197–243. Heller, R., Stanley, D., Yekutieli, D., Rubin, N. & Benjamini, Y. (2006). Cluster-based analysis

Kalisch, M. & Bühlmann, P. (2007). Estimating high-dimensional directed acyclic graphs with

Kaminski, M. (2005). Determination of transmission patterns in multichannel data, *Phil. Trans.*

Kasess, C. H., Stephan, K. E., Weissenbacher, A., Pezawas, L., Moser, E. & Windischberger,

C. (2010). Multi-subject analyses with dynamic causal modeling, *NeuroImage*

the PC-algorithm, *Journal of Machine Learning Research* 8: 613–636.

networks., *Proc Natl Acad SciUSA* 99(12): 7821–7826.

cross-spectral methods, *Econometrica* 37(3): 424–438.

of fMRI data, *NeuroImage* 33(2): 599–608.

*R. Soc. B* 360(1457): 947–952.

49(4): 3065–3074.

approach to structure discovery in Bayesian networks, *MACH LEARN* 50(1): 95–125.

and to related problems, *Journal of the American Statistical Association* 72(358): 320–338.

structural and functional systems., *Nat Rev Neurosci* 10(3): 186–198.

Breakspear, M. (2004). Wavelets and functional magnetic resonance imaging of the

from functional mri data using independent component analysis, *Human Brain*

functional connectivity of the brain, *The Annals of Applied Probability* 9(4): 1021–1057.

under dependency, *Ann. Stat.* 29(4): 1165–1188.

human brain., *Neuroimage* 23 Suppl 1: S234–S249.

*Mapping* 14(3): 140–151.

*ACM (JACM)* 17(1): 51–64.

98(24): 13763–13768.

*Rev. E* 71(5): 056103.

40(1): 35–41.

19(4): 1273–1302.

128: 2224–2239.

Bollen, K. A. (1989). *Structural Equations With Latent Variables*, John Wiley.

identification, *Journal of Mathematical Sociology* 2(1): 113–120.


**20** 

*1,2,4Argentina* 

*3Chile* 

**Event-Related Potential Studies of** 

**Cognitive and Social Neuroscience** 

*2National Scientific and Technical Research Council (CONICET), Buenos Aires, 3Laboratory of Cognitive Neuroscience, Universidad Diego Portales, Santiago,* 

In this chapter, we assess the role of Event-Related Potentials (ERP) in the field of cognitive neuroscience, particularly in the emergent area of social neuroscience. This is new ground that combines approaches from cognitive neuroscience and social psychology, highlighting the multilevel approach to emotional, social and cognitive phenomena, and representing one of the most promising fields of cognitive neuroscience (Adolphs, 2003, 2010; Blakemore, Winston and Frith, 2004; Cunningham and Zelazo, 2007; Decety and Sommerville, 2003; Frith and Frith, 2010; Insel, 2010; Lieberman and Eisenberger, 2009; Miller, 2006; Ochsner, 2004; Rilling and

The technique of ERPs is a precise tool regarding time resolution (on the order of milliseconds). ERPs are useful not only for their excellent temporal resolution but because recent advances (e.g., dense arrays, single trial analysis, source localization algorithms, connectivity and frequency measures, among others) provide multiples sources of brain

First, a definition of ERPs and an explanation about the recordings and features of main components (P1, N1, N170, VPP, EPN, N2, P2, P3, N400, N400-like LPC, LPP, P600, ERN, fERN, CNV, RP; LRP, MP, RAP) are detailed (including a description of their generating sources when available). We then introduce some representative examples of cognitive and social neuroscience: contextual approaches to language, emotions and emotional body language; empathy; and decision-making cognition. All these areas are reviewed, highlighting their relevance for cognitive neuroscience and clinical research (neuropsychiatry and pathophysiology). Finally, important issues, such as sleep research, intracranial ERPs

recordings, source location in dense arrays and co-recordings with fMRI, are discussed.

The technique of ERPs is a precise tool regarding time resolution (on the order of milliseconds) that incorporates the recording of ongoing electrophysiological activity using

Sanfey, 2011; Sanfey, 2007; Singer and Lamm, 2009; Zaki and Ochsner, 2009).

**1. Introduction** 

activity in response to cognitive events.

**2. Event-Related Potentials (ERPs)** 

Agustin Ibanez1,2,3,4, Phil Baker1 and Alvaro Moya1 *1Laboratory of Experimental Psychology and Neuroscience (LPEN),* 

*4Institute of Neuroscience, Favaloro University, Buenos Aires,* 

*Institute of Cognitive Neurology (INECO), Buenos Aires* 

Spirtes, P., Richardson, T., Meek, C., Scheines, R. & Glymour, C. (1998). Using path diagrams as a structural equation modeling tool, *Sociological Methods Research* 27(2): 182–225.


Sporns, O. & Kötter, R. (2004). Motifs in brain networks, *PLoS Biol* 2(11): e369.
