Dear Nacho and SPM'ers
In short:
The more you need the slicetiming (long TR) the more the assumptions for
slicetiming are violated.
The somewhat longer explanation:
Slicetiming relies on the assumption that the acquired signal is
band-limited and can therefore be interpolated based on our samples.
While this is probably true for the BOLD signal (there is not much power
in the BOLD response above 0.1 Hz) it is definitely not true for the
coloured physiological noise (cardiac pulsation 1Hz, respiratory
movement: 0.15Hz). These noise sources gets aliased when you do not
sample them critically (i.e. TR shorter than 500ms for the fundamental
cardiac frequency). This means that slicetiming will go wrong when
trying to estimate what the signal in a given slice would have been had
it been acquired at the same time as the reference slice. And the more
under sampled the cardiac noise is the more incorrect slice-timing will get.
One reason for using interleaved acquisition is to avoid cross-talk
between slices, and moreover some people has also found the interleaved
acquisition to be more temporally stable. The drawback is that if you
want to smooth afterwards you would like the BOLD signal in adjacent
voxels (slices) to be the same, and not delayed by TR/2. Again the
longer the TR the bigger the problem. If you want to smooth anyway you
should not care too much about cross talk anyway, and then you should
just use a flexible basis set and ascending acquisition. If you are
interested in high spatial resolution you would not accept gaps nor
cross-talk, and I would go for the interleaved acquisition and again a
flexible basis set. Another approach would be to use a slice specific
designmatrix like in Worsley's fmristat, but this still doesn't solve
the issue of smoothing like with like.
In SPM99 the smoothing issue could be solved by using smoothed contrast
images for the second level analysis, but with SPM2 (which uses ReML
estimation) this approach is not strictly correct any longer.
Hope this helps
Torben Lund, Danish Research Centre for MR
Ignacio Vallines wrote:
> Hi SPMers,
>
> we are using a Siemens Sonata 1.5T running Syngo... could anyone point
> me out what the advantages/drawbacks of using interleaved aquisition
> are???? of course for fine event related designs where timming is
> crucial.... how reliable is the time correction algorithm in SPM2???..
> what about the influence of short TRs??? we have a new eight channel
> headcoil that can do 20 slices in less that a second... does it make
> any sense with 1.5 T???? I have heard arguments and explanations for
> and against it... what is the feeling from an SPM users perspective????
>
> I wold really appreciatte tips on this!
>
> thanks!
>
> Nacho
>
|