019s correlation and transfer
entropy [5] have been used).
1Academic Radiology, University of Nottingham, Nottingham, United Kingdom
2School of Mathematical Sciences, University of Nottingham, Nottingham, United Kingdom
Introduction There is a growing interest in the investigation of spontaneous brain function (SBF) using fMRI as a potential tool for exploring healthy brain function as well as for development of disease biomarkers.
A pair of features of SBF clearly stands out: Firstly, coherent spontaneous low-frequency (0.01-0.1 Hz) activity fluctuations (SLF) [1], and secondly, SLF coherence for specific large-scale networks of brain regions. We speak about functional connectivity (FC) between these regions.
Use of neural mass model to explore the possible mechanisms of SBF was adopted by Honey [2]. In our study we utilise less computationally demanding model allowing us to observe the behaviour for range of parameter settings. Two phenomena we focus on are FC agreement with underlying anatomical structural connectivity (SC) and frequency of SLF.
Materials and methods We use a connection matrix of macaque cerebral cortex comprising of 47 visual,
sensory and motor areas linked by 505 pathways identified by anatomical tracing studies [3]. We diverge from
Honey by employing a classical neural-mass model for the dynamics, connecting 47 Wilson-Cowan type with
modules by excitatory-to-excitatory coupling employing sigmoidal firing rate function (model implemented in
MATLAB, sampling rate 10 kHz, two different measures of FC, namely Pearson019s correlation and transfer
entropy [5] have been used).
Results We demonstrate that high FC/SC agreement depends critically on parameter settings of the model. In SLF investigation, spectrum of band-limited power signal did indeed show power in low-frequencies but not uniquely in the 0.01-0.1 Hz range as reported in fMRI and LFP studies [1], [4].
Discussion and conclusion The dependence of FC/SC agreement on model parameters outlines important consequences for studies of this type. Still, our model does not so far fully mimic the in vivo observed temporal scale low-frequency fluctuations which to our best knowledge have not been explicitly computationally modelled so far. Our next aim is to analyse the effect of heterogeneity in temporal scales of modules and non-binary connectivity matrix and to implement the between-module transmission delays. The ultimate goal is being able to make testable predictions on individual’s resting state fMRI BOLD signal pattern using individual’s DTI.
Literature cited [1] Fransson, P. 2006. Neuropsychologia 44(14): 2836-2845. [2] Honey, CJ et al. 2007. 104 (24):10240-10245. [3] Kotter, R. 2004. Neuroinformatics 2 (2):127-144. [4] Leopold, DA et al. 2003. Cerebral Cortex 13 (4):422-433. [5] Schreiber, T. 2000. Phys Rev Lett 85 (2):461-464.