There needs to be standardized recording parameters for MRI and fMRI. In looking at studies on area measurement of the corpus callosum I found 7 different techniques in the first 7 articles I looked at...hard for between-study consistency. Combining techniques is probably the best approach for the best results.

# Joel E Alexander

Thanks but what if you are doing brand new research comparing previous in control participants to a brain damaged sample? My idea was to take fMRI before doing the training and then take another fMRI after, using a within-subjects counterbalanced technique... Do you think that will work? How would you standardise the recording parameters?

Thanks, Nick

The longer the time between within subject sessions the more problems. I would personally use the graph theory technique, it seems to pull out the whole brain differences better than ICA or Seed.

Thanks a lot

Hi Nicholas,

so first of all, there will of course be differences. You are mainly looking for what is called 'inter-session variability', and is indeed a hot topic in the neuroimaging research community. Generally, there are number of papers around this topic, see e.g.

http://scholar.google.nl/scholar?hl=en&q=inter+session+variability+fmri&btnG=&as_sdt=1%2C5&as_sdtp=

I am not an fMRI expert, but I know a thing or two about EEG and MEG :) but there you suffer from similar issues. There is loads of literature out there to read about. For EEG and MEG, most literature concerning inter-session variability is on brain-computer interfacing, but there are also some other, more fundamental papers.

As a sidenote, EEG and MEG are different recording techniques, ERPs are a result of the analysis of EEG data. Don't confuse such things ;)

Also, I wouldn't say that graph theory techniques are better than ICA or a seed region analysis.First, it depends on what goal your analysis has, second you have to quantify graph theoretical properties by some measure or technique, and you can also apply graph theory approaches on ICs.

Hi,

all neuroimaging techniques are far from beeing perfect w.r.t retest reliability. For BOLD fMRI activation, please see

http://www.sciencedirect.com/science/article/pii/S1053811912001565

There are some papers that show higher reliability for some EEG parameters (>.90; see e.g. Fallgatter AJ, Aranda DR, Bartsch AJ, Herrmann MJ. Long-term reliability of Electrophysiologic response control parameters. J Clin Neurophysiol. 2002.)

However, the spatial resolution of EEG is of course worse compared to fMRI and that might be one reason for this.

Interestingly, regional "unreliabiliy" can also be a sign of neuroplasticity. Therefore, one should always check whether unreliability is based on "noise" or on some kind of meaningful functional change in neural structures (e.g. training effects!).

Cheers

Michael

Thanks everyone for the answers, it has been really helpful.

Some of the deeper structures (POA, BG, PCC) would be virtually impossible to visualize with MEG (especially on a singe subject level).

The success with your test/re-test variation in fMRI is highly dependent on the functional stimuli - see for example (Morrot et al., j Magn Reson Imaging, 2012) on olfactory stimuli.