> -----Original Message-----
> From: SPM (Statistical Parametric Mapping)
> [mailto:[log in to unmask]] On Behalf Of Sophie Josee Lafaille
> Sent: Friday, 25 November 2005 7:49 AM
> To: [log in to unmask]
> Subject: [SPM] onsets don't match matrix
>
> I'm using SPM2, auditory paradigm (2sec acquisition every 10 seconds).
> no dummy frames, no frames removed during recon.
>
> Each frame is one event. I'm using FIR with a 2 second
> window (default bins set to 1)
If you havent already, have a look at the SPM email archives, maybe do a
search for:
"sparse" or FMRI_T0 for some more info regarding this.
> 2 conditions
>
> word presentation or blank (20 of each)
>
> why does my matrix not match my onsets (onsets start at 0)
>
> Words
> 1 2 4 5 9 11 12 14 19 20 22 24 26 29 30 31 33 34 35 38 Blank
> 0 3 6 7 8 10 13 15 16 17 18 21 23 25 27 28 32 36 37 39
Looks OK to me, but make sure you do have 40 images in your time series and
your durations and TR are not somehow incompatible. See below, but I think
you should set both to 10 secs. This seems to contravene the general rule
that stimulus timing should not be correlated with image acquisition timing
but its difficult to avoid this in sparse designs.
>
>
> am I wrong in concluding that SPM removes my 0 onset for some
> reason? (I
> thought SPM took frame1 for onset 0?) Am I just reading the
> matrix wrong?
Yes you are right, an onset occurring at the beginning of the first image
(frame) in the time series should always be zero. A value of zero does seem
wrong for sparse designs, but since at the very moment the first image
acquisition commences no time has elapsed in your image time series, a value
zero is correct. SPM doesn't remove the first scan or onset though.
For your zero onset above, you are pretending that that the onset of the
very first Blank was coincident with the commencement of the very first
image acquisition. For a sparse design such as yours, the onset of the first
Blank should have really occurred some time during the 8 sec interval
immediately preceeding the first 2 sec image acquistition period (where your
TR = 10secs). So long as you set time bins to 1 then my understanding is
that this neccessary pretence is of no great consequence, but see the FIL's
advice in the email archives.
For most scanners it is wise to entirely remove the first epi images in a
time series to avoid T1 saturation effects and adjust onsets accordingly.
You could check to see whether, for your scanner and sparse design, your
initial 1 or 2 images show very intense values.
>
> how do I remedy this?
>
> also, just as a quick aside, I am using the right technique
> for analysis (FIR)?
> I can't think of another way (ie specifying the HRF in
> nonexistent time?).
You can use the HRF when time bins are set to one, as counter intuitive as
that may seem. Obviously this precludes an event-related analysis, but check
the email archive for expert advice.
>
>
> thanks a billion, me love you long time too ;-)))
>
Thanks very much but not necessary...no really :-) I do have an ulterior
motive for responding though, as I would appreciate any discussion of this
(particularly if my comments above are incorrect). Just seems to be a basic
incompatibility between SPM (or maybe just time series analyses) and sparse
designs, so be interested to hear whether there are any more recent
solutions or advice about this.
Cheers
-Bill
> soph (SPM user since '96)
>
>
>
> --
> Sophie Lafaille, B.Sc.(Hons.), M.Sc.
> Research Officer II, Speech Fluency Laboratory 1059
> Graduate Department of Speech-Language Pathology
> 500 University Avenue, Toronto, Ontario, Canada M5G 1V7
> Ph. 1-(416)-946-8635 Fax 1-(416)-978-1596
> We don't stop playing because we grow old, we grow old
> because we stop playing.
> - Anonymous
> http://to05.endcancer.ca/site/TR?pg=personal&fr_id=1000&px=1208160
>
|