Dear Leon:
I think what is difficult about your request, is that the question you
are asking is not really answerable by DCM. DCM cannot show you
changes over time within a single model (well perhaps it can if you
input a variety of lagged onset times as driving or modulatory inputs
into your model, but this would add so many areas and so much
complication from what you've written below, that it really isn't
testable as a starting point.)
In order to design your DCM you are going to have to make some choices
regarding regions, and decide on your model hypothesis. I suggest
reading through the SPM manual on DCM and Klaas Stephan's 10 Simple
Rules for DCM paper (Neuroimage, 2010) to provide more background.
In general a DCM hypothesis should either be of the form in which 1)
you are trying to decide on the best model- i.e., given a particular
set of regions how they should be hooked up to one another or 2) given
a particular model you are exploring how it responds to different task
contexts, e.g,. complex vs. neutral words.
You need to decide how you are driving the model, i.e. what set of
events or trials cause activity in the model.
You also need to decide how the connections between different regions
are modulated by your task. For example, if producing an emotional vs.
non-emotional responses has some differential effect.
As an off-the-cuff example let's say you want to include all the
regions from the pooled responses: The areas are bilateral: medial
prefrontal, lateral prefrontal, mid-temporal regions, and anterior
insula.
I don't know if you have hypotheses about how these regions should be
hooked together. But if not, it will be difficult to test all the
possible models since there are over 16,000,000 ways to hook up just
the intrinsic connections for this model (assuming each region must be
connected to at least 1 other area). If you then have to model where
the driving and modulatory inputs go, the number of potential models
climbs enormously. So let's assume that within a hemisphere all areas
are hooked to all other areas and between hemispheres the regions are
connected to their homologous counterpart.
Your driving input could be all words, perhaps driving both
mid-temporal regions and your modulatory inputs could be the word
onsets divided into different modulatory inputs by whether the subject
made an emotional or neutral response. If you ran this DCM and
examined the modulatory parameters for emotional vs. neutral
modulatory connections you might be able to find a connection that
responded more to one than the other.
This is just one example of many, many possible choices, and as Klaas
recommends in his paper it is important to start with a specific
hypothesis and then decide how you want to test it.
Regards,
Darren
On Tue, Jun 28, 2011 at 1:52 AM, Leon Petchkovsky
<[log in to unmask]> wrote:
> Dear SPM-ers
>
> this is a general question about how to design a DCM model in SPM 5.
>
> We have recently re-analysed the data from an FMRI study of so-called
> "complexed" /emotional responses to Jung's original Word Association test
> (14 subjects, 100 words, presented at 20 second intervals) and compared
> "complexed" responses aganist neutral ones.
>
> In summary, SPM5 analysis of pooled complexed responses revealed areas of
> bilaterally symmetrical activity in medial prefrontal, lateral
> prefrontal, and mid -temporal regions, and anterior insulae. (L caudate
> was also involved but didn't get above the FWE threshold)
>
> In the first 3 seconds, activity well above the FWE FDR threshold is seen
> in (1) L Broca’s pre Broca’s BA44/45, Z 5.58 (2) the L Middle temporal
> lobe Z 4.96, (3) L SMA/middle cingulate Z 4.97 and (4) L Anterior Insula
> Z 5.11 (probably reflecting L intrahemispheric antero-posterior,
> latero-medial and cortico-mesencephalic processes to do with conflicted
> word searching ).
>
> But there is also homologous activation of all these areas in the R
> hemisphere, albeit more weakly. (R Broca Z 4.85, R Mid Temp Z 4.77, R
> insula Z 4.77), suggesting that interhemispheric negotiations are also
> involved.
>
> However, there is a SMALL area in R prefrontal region at 54,4, 47 (R
> precentral Area 6), which stands out on its own, 16 voxel Z 5.38, p FWE
> corr .001. This is the second strongest response in the entire result list.
> I don’t know what to make of it.
>
> 2 seconds further into the scan, a similar pattern obtains. But there is
> now some brief R supraorbital (43 29 -6) activity (p FWE corr .038, Z
> 4.14) probably to do with the executive effort of the conflictual nature of
> the “complexed’ response. And prominence of medial frontal activity shifts
> from L to R SMA Z 5.28 !
>
> The rest of the 20 second epoch achieves nothing that passes the FWE FDR
> threshold.
>
> So I suppose a model might say that there is some some antero-posterior
> (receptive-expressive verbal) activity; with inter-hemispheric negotiations
> which include SMA and cingulum; and cortico-mesencephalic-limbic
> (Bilateral Anterior Insula) interaction, with R supraorbital (executive
> function) region coming on line a little later.
>
>
>
> But this still doesn’t account for the weird SMALL but very prominent R
> precentral Area 6 region at 54,4, 47 .(Z 5.38)
>
> IWe'd love some help, comments, whatever, on how to design a DCM
> hypothesis, or even if it is possible, given there are so many areas
> involved.
>
>
> cheers and joy
>
> Leon
>
>
>
> --
> A/Prof Leon Petchkovsky
> Psychiatrist, Jungian Analyst
> Past President, ANZSJA
> Director, the Pinniger Clinic, HQ@Robina
> www.pinnigerclinic.com
> 0755809397
> 0420923885
>
--
Darren Gitelman, MD
Northwestern University
710 N. Lake Shore Dr.
Abbott 11th Floor
Chicago, IL 60611
Ph: (312) 908-8614
Fax: (312) 908-5073
|