Dear All
NOTES ON THE YNIC ANALYSIS SUPPORT COLLOQUIUM ON VIRTUAL ELECTRODE ANALYSES IN MEG, WEDNESDAY 13 FEBRUARY 2008.
The current YNiC MEG tools allow us to create beautiful and defensible images showing how the brain's responses to stimuli evolve over time. But beamforming requires the comparison of active windows with passive windows, and the use of active windows less that 200 ms long is not recommended. So while moving windows convey a good initial impression of the progression of neural events, they do not take full advantage of MEG's potential for looking at the detailed timing of those events.
There is another issue with beamformed brain maps, which is that it is not easy to compare results across conditions. You can point out that there is a significant response in a particular brain region in one condition of an experiment but not in another, but it could be that the response just makes it over the threshold in one condition and just falls below threshold in the other. The absolute difference in the magnitude of the two responses (or, to be accurate, the magnitude of the difference between the two active and passive conditions) may be small.
Virtual electrodes offer the potential for asking intelligent questions about the strength and time course of events at specific points in the brain, allowing direct comparisons between conditions and making better use of the temporal resolution of MEG. Piers Cornelissen presented the results of some virtual electrode analyses he and his colleagues have done, comparing responses to different types of stimuli at different regions of interest in the brain. Piers showed results in the form of time-frequency histograms ('Stockwell plots'), separating out evoked and induced components of the total response; also event-related frequency (ERF) plots showing changes in the amplitude of evoked responses over time. He also discussed techniques that he and his colleagues have used for making statistical comparisons across conditions in both types of analysis.
A lengthy and gripping discussion ranged over:
1. The best method to identify regions of interest, 2. How best, having identified a point of interest in a standardised brain, to identify the corresponding points in individual participant brains, and 3. What issues are involved in doing individual and group level analyses, and what would be the best analyses to adopt within YNiC, at least as interim solutions that will allow research groups to make progress with these techniques.
Doing virtual electrode analyses at a group level is a complex business but has strong appeal to a number of researchers. Piers has a working solution which he and Uzma will document and make available. In complement to this documentation, YNiC will document the assumptions and issues associated with Virtual Electrode Analysis at the group level.
It was also clear in the discussion that connectivity measures applied to Virtual Electrode analysis are going to be important to understanding more about the signals recorded in MEG, and more about how functional interactions involving different regions in the brain evolve over time. Deep thought will be given to this topic within YNiC.
Ideas for future fora are welcomed.
Andy Ellis and Michael Simpson