The main research goal of the LHBD is to understand the basic brain functions and relate them to normal and pathological behavior. Apart from being a fascinating intellectual exercise, this will enable us to recognize what makes the brain go astray, and how to protect it or make it right again. The better understanding of the brain functions will help us create neural prosthesis, develop superior brain-machine interfaces and other similar devices for people with disabilities, or even for augmenting the normal brain wherever and whenever it is beneficial.

To achieve this goal, at LHBD we study the brain using many different methods, but with special emphasis on single trial tomographic analysis of MEG data.

Brain function and MEG

Brain function is associated with mass coordinated neuroelectric activity across wide temporal and spatial scales. This activity produces electromagnetic signals that instantaneously (at the speed of light) propagate outside the head and can be measured by electroencephalography (EEG) and magnetoencephalography (MEG). Currently, EEG and MEG are the only non-invasive neuroimaging techniques that measure directly the mass neuronal activity and provide sub-millisecond temporal and millimeter spatial resolution necessary for understanding the entire range of brain functions. See our key findings using MEG.

Brain response variability

Brain responses are very variable. Recording the responses to identical stimuli produces different pattern of EEG / MEG signals, even when all external noise components have been eliminated. The usual way round this problem is to attribute the variability to some uninteresting background activity or "brain noise" and average a large number of signals recorded while identical stimulus was presented. Although this procedure leads to well-defined and largely reproducible response it eliminates activity which although not precisely timelocked to the stimulus, it does nevertheless relate to its processing. If we are to understand brain function, we need to be able to analyse this variability, it has to be analyzed in a quantitative way. To achieve this goal we have developed new ways of looking at variability and demonstrated that the model (signal + noise) that is the foundation of the averaging process is not valid (see Laskaris et al., 2001, 2002, 2003). In summary we have shown that variability of mass electrical activity is not simply an epiphenomenon; it reflects local and global function-related brain dynamics. More ... 

How well can MEG localize?

An open question about the sources of recorded MEG (and EEG) signals is how focal they are and how precisely they can be identified. Averaging can "clean up" some of the extended activity but it also mixes activations from different neural sources that are not precisely timelocked. We have shown that statistical analysis of distributed solutions obtained from MEG signals leads to very precise foci of activation very similar to the ones obtained by fMRI. The difference of course is that the statistical parametric maps (SPMs) derived from MEG signal can map activity in the brain every few milliseconds.

We have addressed the question of how well MEG can localize in a combined MEG and fMRI study, where we have compared the activations in primary visual cortex in the same subjects. We have taken every precaution to optimize each method so that failure to identify common loci within V1 could not be attributed to technical limitations or contributions from extrastriate areas. This result has demonstrated that the early entry to V1, about 40 ms after stimulus onset, was identified with MEG within a few millimeters of the V1 foci identified by fMRI for the same subject and stimulus (see Moradi et. al., 2003). This accuracy was at the limit of the MEG-MRI coregistration accuracy at the time and has lead to the development of new, but tedious methods to improve the coregistration to within one millimeter.

In the follow-up MEG-only study we stimulated 16 different locations in the visual field: at two different eccentricities (4 and 9 degrees) in each quadrant of visual field and on visual meridians (vertical and horizontal). For each subject the same experiment was repeated on three different days, months apart. The results revealed localization precision of within two millimeters throughout visual cortex (see Poghosyan and Ioannides, 2007). Additionally, the results provided the precise timing of early visual areas, and for the first time in humans showed significantly shorter latencies in V1 for peripheral than parafoveal visual field stimulations.

The demonstration of this remarkable spatial localization does not imply that MEG is able to 'see' deep sources with comparable accuracy. It is indeed one of the aims of LHBD to look at the correlation between deep source physiology and what we know from animal and human work carried out during operations. We have already been able to show that deep sources in the brain stem and the cerebellum, and amygdala can be accurately localized (see Ioannides et. al., 2004). More ... (summary and images)