Hi all,
Thank you for posting your paper, Artur - this was helpful. I have a follow-up question re: rationale for using a "balanced template" based on equal numbers of subjects from each group (similar to Routine 6 in Marchewka et al., 2013, HBM) vs. including all subjects to create the custom template (Routine 1). This paper supports the use of the latter (all subjects) over a "balanced template". However, the balanced template took into account three variables (diagnosis, gender, and magnet field strength) and therefore resulted in smaller n's per group included in the template, as opposed to a single variable of diagnosis.
Does anyone have any additional references to inform the choice of whether to use a "balanced template" across diagnostic groups that is based on a single variable of diagnosis?
We are running a VBM analysis in SPM8 (using DARTEL tools) across 5 groups (n=48, n=28, n=16, n=46, and n=16) of children/adolescents. We plan to create a custom template that includes subjects from ALL groups (I.e., the control group AND the 4 patient groups), but need to justify whether to create the template based on all subjects or a balanced subset of subjects across diagnostic groups.
Thank you,
Andrea
On Mon, 1 Jul 2013 16:03:40 +0200, Artur Marchewka <[log in to unmask]> wrote:
>Dear Oscar and Jonathan,
>you may find some answers to your questions in our recent work:
>
>http://onlinelibrary.wiley.com/doi/10.1002/hbm.22297/full
>
>Marchewka, A., Kherif, F., Krueger, G., Grabowska, A., Frackowiak, R.,
>Draganski, B. and The Alzheimer's Disease Neuroimaging Initiative
>(2013), Influence of magnetic field strength and image registration
>strategy on voxel-based morphometry in a study of Alzheimer's disease.
>Hum. Brain Mapp.. doi: 10.1002/hbm.22297
>
>Best regards, Artur
>
>On 01-Jul-13 3:22 PM, Jonathan Peelle wrote:
>> Dear Oskar,
>>
>>> using SPM for VBM analysis of adult healthy/Parkinson/(dementia type AD/DLB) subjects, do you need to create a study-specific template, or can/should you use a generic one, e.g. the IXI template in MNI space that is included in Christian Gaser's VBM toolbox?
>> In principle, I am always more comfortable using a template that is
>> composed of images from subjects similar to those in the current
>> experiment. So if doing a study on healthy adults, the IXI template
>> included in the VBM toolbox would be perfect. For patients, the worry
>> would be that if you have a template based on healthy brains, patient
>> brains will be more difficult to deform in order to match the
>> template. Thus, you have a confound (one group is easier to spatially
>> normalize than another group) that may impact the results.
>>
>> In practice, I don't know how much this affects results, and I am not
>> aware of anyone who has compared this. So, the actual impact on your
>> comparison is hard to know.
>>
>> For myself, I try to always use a template composed of equal numbers
>> of people from each group in question (if the study is unbalanced,
>> this may involve creating a template from a subset of subjects).
>>
>> (Another caveat is that the number of subjects may also play into the
>> accuracy: if you have 10 patients, a template based on these 10 scans
>> may be worse than a template based on 200 healthy brains, simply
>> because of the amount of data available.)
>>
>> Would be interested to hear others' experiences on this however�
>>
>> Hope this helps!
>>
>> Best regards,
>>
>> Jonathan
>>
>
>
>--
>Artur Marchewka Ph.D
>
>Laboratory of Brain Imaging (LOBI)
>Neurobiology Center
>The Nencki Institute of Experimental Biology
>Pasteur 3, 02-093 Warsaw, Poland
>email: [log in to unmask]
>website: http://mruserclub.nencki.gov.pl/
>website: http://lobi.nencki.gov.pl/
>
>
|