Dear Hector,
You could use a full factorial design (one factor with 3 levels) and add
your covariates there (with options "Interactions" and "Centering" set
to "Factor 1"). Then you can test for a group x covariate interaction
with an F-contrast [0 0 0 1 -1 0; 0 0 0 0 1 -1 0].
Best regards,
Guillaume.
On 03/04/17 11:54, Hector Espinos wrote:
> Dear SPM user,
> I am relatively novice in the neuroimaging analysis and I would like to
> have your opinion as experts.
> I'm doing a study with fMRI imaging technique for detecting the Minimal
> Hepatic Encephalopathy (MHE) in early stages.
>
> I have done all the first level analysis and I am doing the second level
> analysis.
> The problem is the following:
>
> I have three groups previously analyzed:
> 17 healthy subjects as a control group
> 22 cirrhotic patients diagnosed with MHE using the PHES (Psychometric
> Hepatic Encephalopathy Score).
> 14 cirrhotic patients showing no symptoms of MHE on phychometric tests.
>
> These three groups done two taks for testing their cognitive
> capabilities:
>
> 1.- Symbol Digit Modalities Test (SDMT). In this task, the subjets
> performed in blocks at two different sppeeds (2 and 2.5 seconds between
> symbols).
> 2.- N-back, where the subjets did a letter identification task performed
> in blocks of two different modalities, O-back and 2-back, with resting
> periods between blocks.
>
> Also, it was registred the activity brain under no specific task
> (resting state).
>
> In the analysis of the first level, the study shows differences in brain
> activation between control subjets and cirrhotic patients. Additionally,
> it could see common areas affected between patients with MHE and no-MHE.
>
> Differences at O-back task involve mainly the lingual cyrus. The 2-back
> task show differences in activation of the caudate nucleus and putamen
> and in some regions of the frontal lobe.
> SDMT shows differences in activation on the middle temporal gyrus.
>
>
> On the other hand, we have a series of measured variables (covariates)
> for the three groups.
>
> Age
> Gender
> Stroop test scores
> Critical Flicker Frequency (CFF)
> Motor coordination (bimanual and visomotor)
> Attention test d2 scores
> Battery Wais scores
> Biochemical parameters (ammonium, cyclic GMP, IL-6, IL-18)
> Markers of oxidative / nitrosative damage (3-Nitrotyrosine,
> hydroxyideoxyguanine, carbonylated proteins, malondialdehyde).
>
>
> My problem is the following:
> I would like to do correlations between the aforementioned variables
> (covariates) and the different groups in order to be able to observe or
> identify significant differences in the responses in order to be able to
> identify which variables or parameters have greater influence and thus
> be able to identify or diagnose the MHE earlier (this is the major goal
> in this study).
>
> At this point, I do not know exactly what would be the best strategy to
> follow. Perform multiple regressions by trying to correlate the
> covariates for each of the groups or to perform a factorial design where
> groups, tasks and covariates are included?.
>
> I would appreciate any possible help to make the most effective design
> that it could provide the maximum information.
>
> Thank you very much in advance.
>
> Hector Espinos-Morato
> INCLIVA Health Research Centre
> Avda. Menendez Pelayo,4
> 46190 (Valencia-Spain)
>
--
Guillaume Flandin, PhD
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
|