Print

---

But what you tell me, I already did it, meaning that i did comparison btw UNI-D and CON, BIP-D vs CON and allDEP vs CON. That was no problem.
And, you are right, i did found a strong effect of age and gender (specially gender) on my data, so i used ANCOVA with age and sex.

; That's sound good.

The uncertainity in my comparison strategy comes when I want to compare UNI-D hyperperfusion and BIP-D hyperperfusion.
But maybe there is no utility to do that... Mmhhh, i'm a bit confused now.

; What do you mean 'hyperperfusion of depression'?
Accordiong to my recent study on major (unipolar) depression using VBM and MRS, there were only degenerative signs. So, I don't think there could be rCBF hyperperperfusion. 
If the patients with major depression of your study were early course of illness, and the patients was medicated with antidepressant, there could be 'adult neurogenesis'. 'adult neurogenesis' could make normal rCBF, but not hyperperfusion compared to normal rCBF.

Sincerely. 

Woo-Suk, Tae
Chuncheon, Korea

2009/11/9 Woo-Suk, Tae <[log in to unmask]>

Hi Christopher Coello

group 1: UNI-D
group 2: BIP-D
group 3: CON

    UNI-D vs CON : 1 0 -1 (UNI-D hyperperfusion compared to normal control)
    BIP-D vs CON : 0 1 -1 (BIP-D hyperperfusion compared to normal control)
    UNI + BIP-D vs CON : 1 1 -2 (hyperperfusion of All Depression compared to normal control)

for the hypoperfusion of depression, just apply the reversed vectors.

But, I think there could be strong age and sex effects on your depression data.
So, I'd like to strongly recommend to use ANCOVA with confounders of age and sex.

Best regards!


Woo-Suk, Tae
Chucheon, Korea

 

--- Original Message ---
From : "Christopher Coello"<[log in to unmask]>
To : [log in to unmask]
Date : 2009/11/09 월요일 오후 6:35:37
Subject : [SPM] SPECT group comparison

Hello SPMers,

I'm being uncertain on the analysis i'm doing these days, so i share it with
you.
We have imaged 3 groups (unipolar depressed (UNI-D), bipolar depressed
(BIP-D) and a control group (CON) ) with SPECT perfusion imaging.
We would like to see if there is a difference between the UNI-D and the
BIP-D perfusion, in contrast with the CON perfusion.
Our idea is to do a multiple regression analysis, with four columns + mean :
UNI-D CON BIP-D CON mean
1 0 0 0 UNI-D group
0 1 0 0 CON group
0 0 1 0 BIP-D group
0 0 0 1 CON group

Then, we used a +1 -1 -1 +1 ((+1 -1) - (+1 -1)) contrast to test an
hyperperfusion (hyper related to control) in UNI-D compared to
hyperperfusion in BIP-D.

Pb : using this contrast, we also test an hypoperfusion in BIP-D conpared to
hypoperfusion in UNI-D.

How could I redo this analysis in order to be able to test these two
hypothesis separately ?

An initial idea was to do a first-level test (UNI-D vs CON and BIP-D vs CON)
and then use the con images (2 images). But i'm uncertain on which test to
do with this two images. Is that a good strategy ?

Thanks in advance for your help,

Christopher Coello,
Postdoctoral Research Fellow
Center for Diagnostic Imaging,
Akershus University Hospital,
Sykehusveien 25, 1478 Lørenskøg
tel : +4745154516
e-mail : [log in to unmask]