Spatio-Temporal Models

We have augmented standard fMRI time series models with spatial priors to provide neuroscientists with more sensitive detection tools. This allows them to find subtle neuronal activations that would otherwise be obscured in experimental noise. The picture below shows bilateral activation of occipital cortex (in red) obtained using a spatial wavelet prior. This provides much better resolution than the standard approach of smoothing the data. My colleague Lee Harrison has used such methods to establish that anterior brain regions integrate information over longer time periods than posterior regions.

• L M Harrison, W Penny, J Daunizeau, and K J Friston. Diffusion-based spatial priors for functional magnetic resonance images. Neuroimage, 41(2):408-23, 2008. PDF

• L M Harrison, W Penny, G Flandin, C C Ruff, N Weiskopf, and K J Friston. Graph-partitioned spatial priors for functional magnetic resonance images. . Neuroimage, 43(4):694-707, 2008. PDF

• G. Flandin and W.D. Penny. Bayesian fMRI data analysis with sparse spatial basis function priors. NeuroImage, 34(3):1108-1125, 2007. PDF

• L. Harrison, W.D. Penny, J. Ashburner, N. Trujillo-Bareto, and K.J. Friston. Diffusion-based spatial priors for imaging. NeuroImage, 38:677-695, 2007.

• W.D. Penny, N. Trujillo-Bareto, and K.J. Friston. Bayesian fMRI time series analysis with spatial priors. NeuroImage, 24(2):350-362, 2005. PDF

Population Receptive Field Models

It is possible to assess the tuning curve of populations of sensory neurons using behavioural measures or fMRI. The images below show amplitude and tuning centres/widths of populations of cells in the human auditory cortex as identified using fMRI (Kumar et al, 2014).

• P Zeidman, E Silson, D Schwarzkopf, C Baker and W Penny(2018) Bayesian population receptive field modelling. NeuroImage, Volume 180, Part A, 173-187. PDF

• S. Kumar and W. Penny (2014). Estimating Neural Response Functions from fMRI. Frontiers in Neuroinformatics, 8:48. PDF

My colleague Peter Zeidman has implemented a Toolbox for estimating PRFs from fMRI data.

Neurovascular Coupling

What aspects of neuronal activity does the BOLD signal (the signal underlying most fMRI studies) reflect ? And how is the BOLD signal related to the MEG or EEG signal ? With my colleague Maria Duarte Rosa we have investigated these issues in the following papers. They provide evidence for the hypothesis that the BOLD signal increases as the Root Mean Square Frequency (RMSF) of the EEG. They also show that BOLD is dependent on both synaptic and spiking activity with the relative contribution changing as underlying cell populations increase their firing rate.

• M. Rosa, J Kilner and W. Penny (2011). Bayesian comparison of neurovascular coupling models using EEG-fMRI. PLoS Computational Biology 7(6) e1002070. PDF

• M Rosa, J. Kilner, F. Blankenburg, O. Josephs and W. Penny. Estimating the transfer function from neuronal activity to BOLD using simultaneous EEG-fMRI. Neuroimage, Jan 15, 2010; 49(4):1496-509. PDF