Dear Dennis,
dennis chen wrote:
> Hello all,
>
> I am fairly new to spm and am using spm96 at the
> moment to analyze spect data. I would like to setup
> an analysis which would compare a group of normal
> patients to both a group of experimental patient scans
> as well as to each individual patient scan within that
> experimental group -- what should be a fairly
> straightforward and simple analysis.
>
> I understand I need to go through steps including
> realignment, coregistration, normalization, smoothing
> and then run the statistics. I am familiar with the
> parameters I need to employ using spect data for
> smoothing and normalization.
>
> I am unclear though as to how I should realign and
> coregister my data (whether to each other, to
> templates, etc. ?) which I assume must be done.
>
My understanding is you have a "one scan per subject" study. In that
case there is no need for "realignment" or "coregistration" which in SPM
jargon denotes the process of registering each scan for a given subject
and modality to each other (realignment) or registering each subjects
scans from different modalities to each other (coregistration). What you
need to do is "normalisation" which refers to spatial normalisation
rather than intensity normalisation. You do that by normalising each
subjects scan to a template, and in your case I assume that the
SPECT.img template that is distributed with SPM will do (provided you
are using a perfusion tracer rather than some fancy receptor ligand).
When you have done that you will have to smooth the data.
>
> Also what form of study design would be most
> appropriate for these questions?
>
For your first question I would suggest using a "Compare populations: 1
scan/subject (two sample t-test) design". It should really be self
explanatory how to use it.
For your second question I would suggest using a "Single subject:
conditions & covariates" design. You would there assign all "normal
patients" to one "condition" and each "experimental patient" to one
condition each. You would then compare each experimental patient to the
normal group through contrast weight vectors of the form [-1 0 0 1]
which example would look for areas with more perfusion in experimental
subject #3 compared to the average of the normal group.
>
> I would appreciate any suggestions/advice. Thank you.
>
> Dennis Chen
Good luck
Jesper
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|