Hello,
This perhaps isn’t the right place to post this, but it seems to be the only online community focused around neuroimaging.
I am having trouble interpreting a Global Field Power plot. The experiment in question simply involved the presentation of face stimuli and object stimuli. For faces, I am seeing a peak deflection at 100ms, and another at around 170ms. However, I also have one around 250 - 300ms. The same thing is seen for objects, but not nearly as much.
I am not sure what the 250 - 300ms peak means exactly, as I am quite new to MEG analysis. I understand a peak around 170ms (m170), but have no idea why there would be a peak at around 250 - 300 ms.
Has anyone conducted a similar experiment that can explain this to me? It might just be the case that I don’t understand GFP plots.
-Ben
Dear Ben,
I don’t know if people in the MEG community use a lot the GFP. This is a tool from the EEG/ERP community, which does not always translate very well to MEG.
You can observe similarities in the early ERP components observed in MEG and EEG, but most of the time the patterns do not match. Because of a combination of instrumental and physiological causes, the timing and shape of the different components can be very different, some standard EEG features like the P300 do not have any clear correspondence in MEG.
The multiplicity of sensor types and sensitivities available in MEG systems makes it difficult to obtain signal shapes and amplitudes that are as reproducible across experiments and subjects as EEG. You’ll get different GFP plots whether you are recordings with axial or planar gradiometers. Therefore it can be difficult to refer to the ERP literature when processing MEG recordings.
These facts lead many MEG researchers to look at different/newer tools to analyze the brain responses, especially with tools exploring the rich frequency dimension of MEG signals, and interpreting the signals as oscillations rather than looking only the peak amplitude of the ERP components (power measures, time-frequency decompositions, connectivity and cross-frequency coupling measures).
As a software engineer, my general recommendation would be not to focus too much on the GFP in MEG, and not to compare your results to the EEG literature.
You can find many examples of visual studies in MEG on our publication page: http://neuroimage.usc.edu/brainstorm/Pub
Specialists of the visual system would be able to help you further with explaining your observations, you could try contacting a few authors directly.
Good luck
Francois
[QUOTE=Francois;11975]Dear Ben,
I don’t know if people in the MEG community use a lot the GFP. This is a tool from the EEG/ERP community, which does not always translate very well to MEG.
You can observe similarities in the early ERP components observed in MEG and EEG, but most of the time the patterns do not match. Because of a combination of instrumental and physiological causes, the timing and shape of the different components can be very different, some standard EEG features like the P300 do not have any clear correspondence in MEG.
The multiplicity of sensor types and sensitivities available in MEG systems makes it difficult to obtain signal shapes and amplitudes that are as reproducible across experiments and subjects as EEG. You’ll get different GFP plots whether you are recordings with axial or planar gradiometers. Therefore it can be difficult to refer to the ERP literature when processing MEG recordings.
These facts lead many MEG researchers to look at different/newer tools to analyze the brain responses, especially with tools exploring the rich frequency dimension of MEG signals, and interpreting the signals as oscillations rather than looking only the peak amplitude of the ERP components (power measures, time-frequency decompositions, connectivity and cross-frequency coupling measures).
As a software engineer, my general recommendation would be not to focus too much on the GFP in MEG, and not to compare your results to the EEG literature.
You can find many examples of visual studies in MEG on our publication page: http://neuroimage.usc.edu/brainstorm/Pub
Specialists of the visual system would be able to help you further with explaining your observations, you could try contacting a few authors directly.
Good luck
Francois[/QUOTE]
Thanks for the answer!
My main aim for this study is to highlight the good temporal resolution of MEG. GFP seemed like an okay way to do that.
Is there any way to calculate a FFT over the GFP in brainstorm?
Thank you!!
No, and this is not recommended.
If you want a synthetic measure of FFT over all the sensors: compute the PSD or FFT over each sensor, and average the power across sensors.