Dear Bo,

>       i have some questions about SPM2 and my fMRI results.
>               1. how can i get hrf curve of active area like fsfast or
>afni show us to judge whether it is a real active relevant to my task?

In voxel-based analyses this is unnecessary.  It may be common practice
to perform post-hoc analyses of regional (ROI) data when using FSFAST
or AFNI but this not a component of SPM.  The judgement of whether an
activation is real has two components. (i) Inference, based on the adjusted
p-value from the SPM and (ii) Quantifying significant effects using the
parameter estimates (or responses associated with those estimates).  To
plot the
HRF for a voxel you can use the plot option in spm_results.  Remember that the
response is of a local average (defined by the smoothing kernel) that can
be regarded as an ROI.  To increase the size of your ROIs simply repeat the
SPM
analysis with a greater degree of smoothing.

In summary, you do a single voxel-based analysis to identify significant
regional effects based on the SPM.  You then quantity those effects by
reporting
the parameters estimated or the fitted responses, at significant voxels.

There is no need to do a whole-brain search and then perform post-hoc
analyses on regions of interest.  If your post-hoc analysis tests for the
same effect as the whole-brain analysis, it is invalid; because you have
biased voxel-selection towards the effect tested for.  If the post-hoc
analysis
tests for anything else, there is no reason to perform the whole-brain
analysis.

In short, the statistical modelling of the data should comprise one model
estimation and inference procedure and entail (i) an inference and (ii) the
reporting of the effect (e.g. contrast of parameter estimates) about which
the inference has been made.


>               2.i made a contrast result A>B in one of my experiments,
>then i found a z>3.5 active area with group analysis(one simple t test). but
>several days before, i found that i can't see the same area in A>rest(rest
>is ~15s just showing a green cross in the whole black screen) but i can find
>it in rest>B. that means at least to task B, i got a 'negative' result,
>didn't i? i found some articles  which discussed this problem and can't
>accept their explanations. I hope get your opinion.

This is a well discussed area.  My opinion is that increases and decreases
in BOLD signal are veridical indices of regional responses.  In PET regionally
specific rCBF decreases are commonplace.  A compelling example can be found in
Frith et al, who show that regional rCBF correlates inversely with the rate
of word presentation.  This is clear evidence for decreased rCBF induced by
each sensory event.  In short, everything is relative; in your case A >= R
> B.

In fMRI, there may be an interesting asymmetry in the rCBF to BOLD
relationship
that 'rectifies' responses, rendering activations more detectable than
de-activations (see Mechelli et al for details).

>       by the way, i found the SPM users increased very fast in Beijing
>(maybe in the whole country) around us. so many people came to our site to
>study or ask questions. you know we had no chance to study SPM
>systematically, just study from one people then teacher to another, so no
>possible to make clear of many important mechanisms of SPM. it is not help
>to us to understand the data and results and maybe we had piled up some
>misunderstanding and wrong steps in process. So is there any way to carry a
>SPM course in China (maybe first time in our site)  like in US or in Europe?

That's an interesting question.  I will discuss it with my colleagues.
I am also copying this to the SPM List to see if there are any volunteers
or other SPM resources people may know of.

With very best wishes,

Karl


Frith CD, Friston KJ. The role of the thalamus in "top down" modulation of
attention to sound. NeuroImage. 1996 Dec;4(3 Pt 1):210-5.

Mechelli A, Price CJ, Friston KJ.  Nonlinear coupling between evoked rCBF
and BOLD signals: a simulation study of hemodynamic responses.
Neuroimage. 2001 Oct;14(4):862-72.