Extracting discriminant information from neuroimages: A multivariate computational framework to analyze the whole human brain

Nenhuma Miniatura disponível
Citações na Scopus
Tipo de produção
Capítulo de livro
Data de publicação
Thomaz C.E.
Leao R.D.
Sato J.R.
Busatto G.F.
Handbook of Neuropsychiatry Research
Título da Revista
ISSN da Revista
Título de Volume
Texto completo (DOI)
© 2010 Nova Science Publishers, Inc.With the increasing anatomical resolution of the neuroimaging methods, clinicians are challenged nowadays, more than ever before, with the problem of detecting and interpreting statistically significant changes on neuroimages that are often distributed and involve simultaneously several structures of the human brain. In this chapter, we describe a general multivariate linear framework that analyses all the data simultaneously rather than segmented versions separately or feature-by-feature. This approach has been specially designed for extracting discriminative information from high dimensional data, dealing with the problem of small sample sizes, and it has been successfully applied in MR imaging analysis of the human brain. The multivariate linear framework is not restricted to any particular set of features and describes a simple and straightforward way of explaining multivariate changes of the whole brain on the original MR image domain, giving results that are statistically relevant to be further validated and interpreted by clinicians.