Modelling functional brain architecture. (360G-Wellcome-056750_Z_99_C)

£94,552

The aim of the proposed work is to create models that enable useful and informed inferences about brain function based upon whole-brain electromagnetic and hemodynamic responses. These models are important because they define the nature of the inferences made. Neuroimaging with electroencephalography (EEG), magnetoencephalography (MEG), and functional magnetic resonance imaging (fMRI) has an established role in nearly every aspect of cognitive, systems and clinical neuroscience. The data analysis procedures are now relatively sophisticated and ensure valid inferences. However, the statistical models employed are rather simple-minded and have little connection to neurophysiological procecesses, or the principles that might underlie brain function. Conventional models of fMRI are a little more sophisticated than most and rest on linear convolution modes of how changes in neuronal activity are expressed hemodynamically. However, even infMRI, interactions among different neuronal populations or cortical areas are precluded. This is important because the conceptual and biological validity of any forward model, of observed brain responses, places fundamental constraints on the validity and usefulness of inferences about that model. The programme of work described below is an attempt to finesse current models and ground them inneurophysiology and conceptual frameworks derived from machine learning. The hope is that inferences about the parameters of these models are empowered because the parameters have an explicit neurophysiological or mechanistic meaning. If successful, this work will facilitate two key ways of integrating theoretical ideas about large-scale brain function and experimental observations. The first rests on using brain responses to estimate physiologically meaningful parameters of neuronal architectures. The second approach uses data to disambiguate among competing models that embody key theoretical distinctions, formulated in terms of neurophysiology or machine learning; it is a relatively simple matter to identify the most likely model, given the data, using Bayesian model selection.

Where is this data from?

This data was originally published by The Wellcome Trust. If you see something about your organisation or the funding it has received on this page that doesn't look right you can submit a grantee amendment request. You can hover over codes from standard codelists to see the user-friendly name provided by 360Giving.

Grant Details

Amount Awarded 94552
Applicant Surname Friston
Approval Committee Scientific Committee
Award Date 2008-09-16T00:00:00+00:00
Financial Year 2007/08
Grant Programme: Title Principal Research Fellowship (New)
Internal ID 056750/Z/99/C
Lead Applicant Prof Karl Friston
Partnership Value 94552
Planned Dates: End Date 2009-09-30T00:00:00+00:00
Planned Dates: Start Date 2008-10-01T00:00:00+00:00
Recipient Org: Country United Kingdom
Region Greater London
Sponsor(s) Prof Raymond Dolan