Dear Eric
Thanks for this important remark. I am not an engineer, but I have checked
this point when we first begun to perform fast imaging for fMRI. More
technical comments will be welcome.
I have a comment about a small part of one of Geraint's recent replies
which regarded low frequency fMRI noise:
> change is very low. Unfortunately many confounding signals of little
> interest (physiologically aliased signals, drifts in scanner gain and so
> on) also change at low frequencies.
Though I certainly agree that there is more noise at lower frequencies,
I have not seen any data to support their explanation in terms of either
a) drifts in scanner gain
or
b) physiologically aliased signals.
Though both of these seem reasonable a priori as explanations for low
frequency noise, I have not found any evidence to support either in my own
data or in the literature. In fact there is even some evidence to refute
them.
In terms of evidence against model a), a drift in scanner gain would
predict a correlation between the whole brain signal and variability in
signal across space. I looked at this and saw no such correlation (has
anyone else looked at this?).
In fact, for scanner noise (defined here as seen on phantoms), I would
rather talk of 2 types of noise :
1- The first one concern the temporal evolution of the whole phantom (==
global intensity) : In our scan (Brucker 2T upgraded for EPI with an
adapted gradient coils fitted inside the scan) we haven't found a clearly
structured noise (no peek frequency), but rather a white noise (sure this
is true only after the steady state has been reached). Is this what you
meant by 'correlation between the whole brain signal' ?
2- The second concern only certain regions : most of it is on the
interfaces of the object (+++ in regions with susceptibly artifacts) or its
projections by ghosts. Its frequency is ~ 0.0038 Hz and is intensity may
reach ~ 0.7 % (min to max). However those results are probably highly re
lated to the imaging parameters (we haven't check that, has anybody done it
?). Due to this distribution, we interpreted it as gradient instability.
Model b) would predict no (or at least a decrease in) low frequency noise
in non-physiological scanned objects. But, low frequency noise has been
observed in water phantoms that looks very similar to that seen in human
subjects. Furthermore, cardiac and respiratory rhythms have a strong
periodic component that would therefore not alias into a 1/f noise
spectrum.
I don't know if the authors of 'localization of cardiac-induced signal
change in fMRI' are SPM listers, but it would be interesting that they give
us the spectra of there observed cardiac noise.
I raise these issues because it would be wonderful if an engineer or MRI
physicist could do some research into understanding the low frequency
components of fMRI noise better. Or perhaps, some has already been done,
and it just has not filtered its way into the mainstream neuroimaging
literature?
Hope that there will be comments on this issue ...
Sincerely
Jack
________________________________________________________________
| Jack Foucher Universite Louis Pasteur |
| Institut de Physique Biologique UPRES-A 7004 du CNRS |
| 4 rue Kirschleger Tel: 33 (0)3 88 77 89 90 |
| 67085 STRASBOURG Fax: 33 (0)3 88 37 14 97 |
| France |
| Faster E-mail: [log in to unmask] |
| Other [log in to unmask] |
|_______________________________________________________________ |
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|