No. 1025

Title:

Spatio-temporal correlations in fMRI time series: the whitening approach

Author(s):

Bosch-Bayard J.(Cuban Neuroscience Center);
Riera-Diaz J.J(IDAC, Tohoku Univ.);
Biscay-Lirio　R. (Institute of Cybernetics, Mathematics and Physics);
Wong K.F.K.(the Institute of Statistical Mathematics);
Galka A.(the Institute of Experimental and Applied Physics, Kiel Univ.);
Yamashita O.(ATR Computational Neuroscience Labs, Kyoto);
Sadato O.(National Institute for Physiological Science, Okazaki);
Kawashima R. (IDAC, Tohoku Univ.);
Miwakeichi F. (Chiba Univ,);
Valdes-Sosa P.(Cuban Neuroscience Center);
Ozaki T. (the Institute of Statistical Mathematics)

Key words:

ｆMRI, time series, NN-ARx, causality, AIC, connectivity, whitening, innovations.

Abstract:

For the purpose of statistical characterization of the spatio-temporal correlation structure of high-dimensional fMRI time series we propose an innovation approach, based on whitening the data by nearest-neighbors autoregressive modeling with external inputs (NN-ARx). It is demonstrated that the whitening step is suitable for elucidating the significant instantaneous spatial dependences between remote voxels. Potential and limitations of characterizing causality by autoregressive modeling are discussed in the context of the study of brain connectivity structure. Results for the analysis of fMRI data recorded during visual and motor stimulus experiment are also shown.