Computational Neuroscience

Challenges and Implications for Brazilian Education

Raimundo José Macário Costa

, Luís Alfredo Vidal de Carvalho

, Emilio Sánchez Miguel

Renata Mousinho

, Renato Cerceau

2,5

, Lizete Pontes Macário Costa

, Jorge Zavaleta

Laci Mary Barbosa Manhães

and Sérgio Manuel Serra da Cruz

1,6,7

Universidade Federal Rural do Rio de Janeiro (UFRRJ), Rio de Janeiro, Brazil

Universidade Federal do Rio de Janeiro (UFRJ), Rio de Janeiro, Brazil

Salamanca University (U.S.A.L), Salamanca, Spain

Rio de Janeiro State University (UERJ), Rio de Janeiro, Brazil

National Regulatory Agency for Private Health Insurance and Plans (ANS), Rio de Janeiro, Brazil

Programa de Educação Tutorial (PET-SI/UFRRJ), Rio de Janeiro, Brazil

Programa de Pós- Graduação em Modelagem Matemática e Computacional (PPGMMC/UFRRJ), Rio de Janeiro, Brazil

Keywords: Education, Neuroscience, Computer Science, Databases, Artificial Intelligence, Cognitive Science.

Abstract: Understanding the core function of the brain is one the major challenges of our times. In the areas of

neuroscience and education, several new studies try to correlate the learning difficulties faced by children

and youth with behavioral and social problems. This work aims to present the challenges and opportunities

of computational neuroscience research, with the aim of detecting people with learning disorders. We

present a line of investigation based on the key areas: neuroscience, cognitive sciences and computer

science, which considers young people between nine and eighteen years of age, with or without a learning

disorder. The adoption of neural networks reveals consistency in dealing with pattern recognition problems

and they are shown to be effective for early detection in patients with these disorders. We argue that

computational neuroscience can be used for identifying and analyzing young Brazilian people with several

cognitive disorders.

1 INTRODUCTION

Understanding brain function remains one of the

major challenges in the scientific community for the

twenty-first century (Abbot 2013). Research on the

subject has been growing exponentially since the

1960s. Neuroscience is a research field that has also

been growing significantly, beginning from the

1980s (Figure 1). Neuroscience aims to study and

analyze the central nervous system (CNS) of humans

and animals, their functions, particular format,

physiology, and injuries or pathologies. This area

has achieved important advances that have been

responsible for positive effects on the quality of life

of patients suffering from, for example, multiple

sclerosis, Alzheimer's disease, Parkinson's disease,

and other diseases related to the CNS (Lent 2001).

However, despite extensive investment in the area,

much remains to be done, especially in relation to

understanding the binding mechanisms between

brain structures and functions at a microscopic level

for cognitive and behavioral processes (Markram

2013).

In the scientific area, the early 90s was

announced as the "decade of the brain". This

denomination originated in the U.S.A. and sought to

encourage the identification of normal

neuropsycobiological processes and related

disorders. Thus, together with significant advances

in computer science and the spread of the Internet,

the area of computational neuroscience flourished

(Schwartz 1990). Since then, works have been

developed and new strategies have been sought for

the development of realistic mathematical and

computer models to simulate the brain.

Most recently in April 2013, major investigative

projects the BRAIN Initiative (NIH 2014) and the

Human Brain Project (HBP 2014) were presented

again, in the U.S.A. and Europe, respectively. These

initiatives presented demands that propose to

revolutionize the understanding of the functioning of

the mysteries of the human brain. Thus, in order to

436

José Macário Costa R., Alfredo Vidal de Carvalho L., Sánchez Miguel E., Mousinho R., Cerceau R., Macário Costa L., Zavaleta J., Manhães L. and

Serra da Cruz S..

Computational Neuroscience - Challenges and Implications for Brazilian Education.

DOI: 10.5220/0005481004360441

In Proceedings of the 7th International Conference on Computer Supported Education (CSEDU-2015), pages 436-441

ISBN: 978-989-758-107-6

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

accelerate the development of new technologies that

will allow researchers and scientists to obtain

dynamic images of the brain in action, showing how

brain cells and complex neural circuits interact at the

speed of thought, thereby expanding knowledge

about how we think, learn, and remember.

Figure 1: Quantitative comparison of articles, published in

the PubMed database, that include the terms "brain",

"education" and "neuroscience" in the title (January,

2015). (Y-axis: Number of published articles X-axis: Year

of publication).

The structure to accomplish the BRAIN Initiative

includes private companies, research centers, and

government agencies, as well as a wide range of

experts ranging from physicians, neuroscientists,

nanoscientists to engineers and computer scientists

as well. The last ones working in the fields of

artificial intelligence, databases, high performance

computing, big data, e-science, games, robotics,

sensors, and social networks, among other (Zhong

2012; NIH 2014; HBP 2014).

In Brazil, significant efforts developed by

neuroscientists is recorded for the development of

knowledge related to the brain and the "Brazilian

Brain Industry". Among the various research

centers, we can mention the work developed at the

International Neuroscience Institute of Natal

Edmond and Lily Safra (IINN-ELS) and the Institute

of Biomedical Sciences of UFRJ (ICB-UFRJ).

Despite these great efforts and the synergy between

computational neuroscience and education still

needs to go further. Novel studies that correlate the

learning disorders faced by children and young

people with various computational techniques ought

to be developed. Such disorders have deep social

relevance in the educational area, for example,

where they may have an effect on truancy,

functional illiteracy, and repeated failures, as well as

the self-esteem of individuals.

The goal of this paper is to report research

directions and highlight the challenges related to the

adoption of computational neuroscience to enhance

the quality Brazilian education. This work also

presents our ongoing computational approaches

related to the classification of patients with dyslexia,

one of the learning disorders that has aroused

interest among Brazilian researchers, health

professionals, and schools teachers.

This paper is organized as follows. Section 2

persent the current challenges of Computational

neuroscience. Section 3 present the researche

opportunities. Section 4 present the researches that

are being conducted by our research team. Finally,

section 5 concludes the paper.

2 CHALLENGES

Computational neuroscience can be used as a toolset

for building intelligent computational systems that

are capable of processing and analyzing large

volumes of (structured and semi-structured)

educational data. Computational neuroscience can

be used to elaborate educational games, or even

developing mobile applications targeted at

diagnostic support and tracking of learning disorders

(Zavaleta et al., 2012).

Computational neuroscience is essentially

interdisciplinary and rests on key three pillars:

neuroscience (the areas of medicine and biological

sciences); cognitive science (psychology); and

Figure 2: Three pillars in computational neuroscience and

prospects for technological investigations supported by

computer science.

ComputationalNeuroscience-ChallengesandImplicationsforBrazilianEducation

437

computer science (artificial intelligence, databases,

e-science, provenance, big data, high performance

computing (HPC), cloud computing, internet of

things, etc.) (Figure 2). However, one of the major

difficulties of computational neuroscience is to

model (mathematically and computationally) a

learning disorder, identify the most relevant

variables and data, transcribe them for technological

solutions, and determine if the computational results

are significant and valid according to medical,

ethical, and educational aspects.

To achieve a technological level compatible with

the multifaceted demands of brain studies in the

twenty-first century, we must consider the structure

of a research agenda that is focused on new models

and the incorporation of (data intensive)

computational techniques traditional to e-science

(Hey et al. 2009) for the development of applications

in computational neuroscience. In this case, it is

possible to advance in the following areas:

a) Development of intelligent and predictive

systems based on artificial intelligence

techniques (Macário Costa et al. 2007, 2008,

2009, 2010, 2011, 2013; Zavaleta et al. 2012)

capable of handling large volumes of data;

b) Use of distributed computational environ-ments

for high-performance processing of

intelligent systems, to support in silico

simulations and experiments based on

scientific workflows (Deelman et al. 2009) of

simulations of brain models (Abbot 2013;

Kubilius, 2014; NIH 2014; HBP 2014).

c) Adoption of management techniques for large

volumes of semi-structured biological data

and processing typically used for big data

(Davison 2010; Berman 2011; Abbott 2013)

— computational neuroscience projects are

gravitating towards mapping increasingly

larger and more complex brain models that

use sensors and data in different formats

(Zhong et al. 2011; Zhong 2012);

d) Incorporation of provenance descriptors (Cruz

et al. 2009, 2012) and data management and

techniques to increase the reproducibility and

reliability of studies in computational

neuroscience (Chen, Zhong, and Liang 2012;

Ciccarese et al. 2013) — these activities tend

to be conducted by interdisciplinary research

teams that are geographically and temporally

dispersed (Chen and Zhong 2013);

The knowledge acquired in the area of

neuroscience can be associated with novel tools and

computational techniques for improving the

opportunities to act on learning disorders.

Dyslexia can be characterized by a failure in the

acquisition and/or development of scholastic skills.

Dyslexia is a learning disorder that affects 3–7% of

the school-age population, and it highlights other

disorders including severe delays in reading, writing,

and spelling, as well as inversion of symbols. This

work is focused in dyslexia due to its singularity and

limited nature of the phonological deficit (Shaywitz

and Shaywitz 1999; Mousinho 2003). Detailed

understanding of the correlations between genetic

variations, brain dysfunction, and cognitive

difficulties is a great challenge in dyslexia research

(Giraud and Ramus 2013) such investigation require

large computing efforts due to the large amount of

data. Currently, an evaluation of a dyslexic

individual takes on average two months until the

establishment of a diagnosis by a qualified team.

Usually, the students referred to the health service

do not have dyslexia, but learning problems of

different orders. Here it we state that the

establishment of new tracking systems, based on

computational neuroscience, will possibly reduce the

queues and speed up access to diagnosis, thus

offering chances of intervention to more children

and young people in a short time period, in a more

opportune, efficient, and socially just manner.

3 IMPLICATIONS

Investigations directed at the tracking of children

and young people with dyslexia at school age will

contribute to establishing evaluations related to the

timely identification and referral of people with

learning disorders in schools, whether public or

private.

Mousinho (2011) states that it is almost a routine

situation to find students with reading problems. In

most schools, classes are composed of children who

are more or less adept at reading. Among the less

skilled children, there are still those that stand out.

Despite their efforts, reading tasks for these children

becomes a laborious and even painful activity. The

effort put into this action is so big that it may even

hinder the pleasure of reading.

Children with more impaired reading tend to

increase the gap between themselves and their peers,

which may increase the incidence of undesirable

consequences such as the loss of enjoyment of

reading, low performance in other disciplines that

depend on reading, and the development of low self-

esteem by the child (Mousinho 2011).

CSEDU2015-7thInternationalConferenceonComputerSupportedEducation

438

Early identification of children with reading and

learning difficulties has become a priority due to the

possibility of eliminating or reducing the detriment

to the children’s scholastic and social progress.

Thus, the adoption of disruptive computational

strategies based on computational neuroscience can

make all the difference in this identification.

One of the most common computational

strategies of computational neuroscience is the use

of neural networks. Neural networks are

mathematical models that simulate biological neural

structures and have computing ability acquired

through learning and generalization. We believe that

the biggest challenge is in gathering all the different

kinds of information (variables) into databases and

developing new mechanisms for analysis and

detection of dyslexic patients.

Therefore, our research efforts were developed

in order to detect which children and adolescents

were at risk of dyslexia. The prior versions of the

intelligent system developed by Macário Costa et al.

(2007, 2008, 2009, 2010, 2011, 2013) is expanding

its database in order to further consolidate the

accuracy of the algorithms. The algorithms were

used to extract the useful patterns from the data

collected during interviews done with the person

responsible for the child who is enrolled in the

school, in the search for a desired pattern to identify

individuals who have the learning difficulties.

In order to support the diagnosis of specific

learning disorders, Macário Costa et al. (2011, 2012,

2013) resent the implementation of a multi-layer

perceptron neural network to probabilistically

classify youth and adult patients with dyslexia.

4 RESEARCH GOALS

This paper aims to discuss the challenges and open

opportunities to develop novel computational

solutions that can contribute as support tools for

specialists to detect and diagnose, in advance,

individuals with learning disorders in our country.

The computational solutions we highlight can be

applied directly by teachers in elementary and

primary schools. We advocate that there is a need to

establish partnerships with public and private

institutions, governments, and research groups, that

can cooperate and provide computational

infrastructure capable of storing and processing, in a

distributed manner, large amounts of semi-structured

data. Besides, it is also important to offer a

commitment to involve their teams in the schools in

order to track the individuals who have learning

difficulties and/or are at risk of dyslexia or other

learning disorders.

The goals of our research is inherently

multidisciplinary and should be performed by people

with different skills (Macário Costa et al, 2014). It is

divided into three phases. The first phase, which has

already been performed, consisted of outlining the

problem and designing the tools and the neural

networks, followed by the construction of the

intelligent system and its database, and then

performing the testing and evaluation of the models.

The second phase, which is in progress, consists

of the survey of new demands necessary for the

scalability of the solution, enabling it to incorporate

the new technological artifacts (discussed above)

that can process large volumes of data in high-

performance distributed environments. In this phase

there is also the mapping of new collaborators and

schools. Without a combination of technologies, we

cannot answer the basic questions such as: How do

we detect — at an early stage, at a low operating

cost, and at high effectiveness — young people at

risk of dyslexia or other learning disorders? In a

country like Brazil, with its continental dimensions

and its historical social inequalities, obtaining

answers to questions of this nature could be a

strategic advantage for the country's future.

The third and final phase is crucial for the

project’s success. It will be more widespread and

will be operational. It will be directed toward the

massive and early detection approaches in the school

environment in order to detect the children with

reading and learning difficulties. This phase requires

accurate, reliable, and quick analysis and processing

of data to assist in the medical diagnosis, and to

establish early referral of these individuals to

specialists.

5 CONCLUSIONS

This research allowed us to develop an intelligent

computational system, using artificial intelligence

techniques for the detection of individuals between 9

and 18 years of age who have learning difficulties

(dyslexia). The intelligent computational system is

composed of several modules, some are completed

and the others are being developed.

The neural network (NN) module allowed the

development of a method to support the process for

identifying people and children with dyslexia and

other learning difficulties. A database was developed

for the collection of data used by the NN module.

This module has been completed and the others are

ComputationalNeuroscience-ChallengesandImplicationsforBrazilianEducation

439

in the development and testing phase.

The individuals identified in the NN module pass

to the Response to Intervention (RTI) pyramidal

module which consists of three layers of evidence-

based interventions for promoting the social,

emotional, and behavioral development of children.

Each layer uses fuzzy logic to assign degrees of

learning difficulty to the individuals and determine

the most appropriate computational intervention for

each layer of the RTI model. Each layer of the RTI

model will consist of a set of computational

intervention methodologies (e.g., games) activated

by the degree of difficulty for each individual.

The proposed approach can be used innovatively

as a support tool for the diagnosis of dyslexia and

other learning difficulties. Further details and

detailed descriptions can be found at our previous

work (Macário Costa et al., 2014).

ACKNOWLEDGEMENTS

The authors R.J.M. Costa and R. Mousinho thank

the team of the ELO Project (Department of Speech

Pathology, and Faculty of Medicine at UFRJ) and

the Delindo Couto Neurology Institute of UFRJ and

USAL. J. Zavaleta thanks CAPES for the financial

support received. S.M.S. Cruz thanks CNPq,

CYTED (Programa Iberoamericano De Ciencia Y

Tecnologia Para El Desarrollo - grant P514RT0013),

FAPERJ (grants E-26/112.588/2012 and E-

26/110.928/2013), MEC/SeSU and PET-SI/UFRRJ

for the financial support of the research.

REFERENCES

Abbott, A. (2013) “Neuroscience: Solving the brain”, In:

Nature 499, pages. 272–274.

Berman JJ, (2009) “Principles of Big Data: Preparing,

Sharing, and Analyzing Complex Information”

Morgan Kaufmann; 1st Edition.

Ciccarese, P., Soiland-Reyes, S., Belhajjame, K., Gray,

A.J.G, Goble, C., Clark, T. (2013) “PAV ontology:

provenance, authoring and versioning”, Journal of

Biomedical Semantics 2013, 4:37.

Chen J. H., Zhong, N. (2013) “Toward the Data-Brain

driven systematic brain data analysis”. IEEE

Transactions on Systems, Man, and Cybernetics:

Systems, 43(1), pages. 222-228.

Chen J. H., Zhong, N., Liang, P. P. (2012) “Data-Brain

driven systematic human brain data analysis: A case

study in numerical inductive reasoning centric

investigation” Cognitive Systems Research, Elsevier,

vol. 15-16, pages. 17-32.

Cruz, S.M.S, Campos, M. L M., Mattoso, M. (2009)

“Towards a Taxonomy of Provenance in Scientific

Workflow Management Systems”. SERVICES I 2009:

259-266.

Cruz, S.M.S, Campos, M. L M., Mattoso, M. (2012) “A

Foundational Ontology to Support Scientific

Experiments”. ONTOBRAS-MOST 2012: 144-155.

Davison, A.P. (2010) “Challenges and solutions in

replicability and provenance tracking for simulation

projects”. BMC Neuroscience 2010, 11(Suppl 1):P76.

Deelman E, Gannon D, Shields M, Taylor I (2009)

“Workflows and e-Science: An overview of workflow

system features and capabilities”, Future Generation

Computer Systems 25(5):528–540.

Giraud A.L., Ramus, F. (2013) “Neurogenetics and

auditory processing in developmental dyslexia”. Curr.

Opin. Neurobiol. 23:37–42.

HBP (2014) “Human Brain Project”.

https://www.humanbrainproject.eu/

Hey, T., Tansley, S., Tolle, K (2009) “The Fourth

Paradigm: Data-Intensive Scientific Discovery”

Microsoft Press. 1st Edition.

Kubilius, J. (2013) “A framework for streamlining

research workflow in neuroscience and psychology”.

Front. Neuroinform. 7: 52.

Lent, R. (2001) “Cem Bilhões de Neurônios – Conceitos

Fundamentais de Neurociência”. SP. Ed. Atheneu.

Macário Costa, R. J., Pará, T. S., Caloba, L. P., Carvalho,

L. A. V. (2007) “Classificação de pacientes com

transtorno de dislexia usando redes neurais artificiais”.

In: XXX CNMAC, Florianópolis.

Macário Costa, R. J., Mousinho, R., Carvalho, L. A. V.

(2008) “Redes Neuronais: um instrumento no rastreio

(screening) de pessoas com risco de transtorno

específico de leitura”. In: I Congresso Ibro/Larc de

Neurociências da América Latina, Caribe e Península

Ibérica, 2008, Búzios, RJ.

Macário Costa, R. J., Mousinho, R., Vidal, L. A. (2009)

“Abordagem Computacional no Screening da Dislexia

e do TDAH”. In: XXXII CNMAC, Cuiabá.

Macário Costa, R. J., Cruz, S. M. S., Vidal, L. A.,

Mousinho, R, Tosta, F. O. (2009) “Redes neuronais e

transtornos de aprendizagem: rastreio de pessoas com

dislexia”. In: SBIE. v. 20. p. 1-10. Florianopólis.

Macário Costa, R. J., Mousinho, R.; Vidal, L. A. (2009)

“Dislexia e Inteligência Computacional: Um sistema

para rastrear (Screening) pessoas com sinais de

transtorno de leitura”. In: 2o. Congresso Internacional

de Dislexia, São Paulo.

Macário Costa, R. J. (2011) “Uma Estratégia

computacional na detecção da dislexia”. Rio de

Janeiro: Tese – UFRJ/COPPE.

Macário Costa, R. J. Cruz, S. M. S., Zavaleta, J, Carvalho,

L. A. V.; Mousinho, Renata (2011) “Abordagem

tecnológica para rastreio de pessoas com dislexia”.

Tecer (Belo Horizonte), v. 4, p. 41-53.

Macário Costa, R. J., Zavaleta, J., Cruz, S. M. S., Cerceau,

R., Mousinho, R., Carvalho, L. A. V. (2013) “A

Computational Approach for Screening Dyslexia”. In:

CBMS 2013, 2013, Porto. 26

IEEE International

CSEDU2015-7thInternationalConferenceonComputerSupportedEducation

440

Symposium on Computer-Based Medical Systems.

Macário Costa, R. J., Carvalho, L. A. V, Miguel, E.

Mousinho, R S., Cerceau, R., Macario Costa L. P.,

Cruz, S. M. S. (2014) “Desafios e Oportunidades em

Neurociência Computacional na Educação Brasileira”.

In: Grandes Desafios da Computação no Brasil 3a

Edição. Porto Alegre. Sociedade Brasileira da

Computação. SBC. (to be published).

Markram, H. (2013) “Seven Challenges in Neuroscience”,

In: Functional Neurobiology 28(3) 145-151.

Mousinho, R. (2003) “Desenvolvimento da Leitura,

Escrita e seus Transtornos. In: Goldfeld, M. Fund. em

Fonoaudiologia”. Guanabara Koogan. 2a edição.

Mousinho, R. et al. (2011) “Dislexia – Novos temas,

novas perspectivas. Wak Editora. Rio de Janeiro.

NIH (2014) “BRAIN 2025: A Scientific Vision”,

http://www.nih.gov/science/brain/2025/.

Schwartz, E. (1990) “Computational Neuroscience”, MIT

Press, 1st edition.

Shaywitz, S.E. Shaywitz, B.A. (1999) “Dyslexia In:

Swaiman KF, Ashwal S. Pediatric Neurology -

Principal e Practice”, Connecticut. Ed. Mosby.

Zavaleta, J. et al. (2012) “DysDTool: Uma Ferramenta

Inteligente para Avaliação e Intervenção no Apoio ao

Diagnóstico da Dislexia”. In: CSBC- XII Workshop

de Informática Médica.

Zhong, N. et al. (2011) “Brain Informatics”. IEEE

Intelligent Systems. September, pages 16-20.

Zhong, N (2012) “Research Issues and Challenges on

Brain Informatics Towards Computing & Intelligence

in the Big Data Era”.

pakdd2014.pakdd.org/tutorial3.pdf.

ComputationalNeuroscience-ChallengesandImplicationsforBrazilianEducation

441