Dear Liang,
On Mon, 9 Dec 2013 22:14:48 +0800, Liang Wang <[log in to unmask]> wrote:
>Dear Christian,
>
>Thanks a lot for your kind clarification. I have a following question about
>the longitudinal VBM model design for your help (for example, three time
>points for two subjects and two time points for the last one). Following
>the VBM manual, I created a design matrix which had four columns as follows:
>
>1.33 1.33 0 0
>1.03 0 1.03 0
>0.86 0 0 0.86
>1.33 1.33 0 0
>1.03 0 1.03 0
>0.86 0 0 0.86
>1.33 1.33 0 0
>1.03 0 1.03 0
Uups, I can't imagine that this is mentioned in the VBM manual. The design you have created below is the right one. After creating the design there should be always "ones" if you have factors that model the time points and the subjects.
Best,
Christian
>
>The 1st column seemed reflecting a fix decreased effects across the
>subjects, but I don't know why it should be setup like this way (??) And I
>have no idea where this specific number came from. Could you please explain
>in little more details? Thanks.
>
>For such model design, I was thinking to use multiple regression to
>implement. The model would look like:
>1 0 0 1 0 0
>1 0 0 0 1 0
>1 0 0 0 0 1
>0 1 0 1 0 0
>0 1 0 0 1 0
>0 1 0 0 0 1
>0 0 1 1 0 0
>0 0 1 0 1 0
>
>Compared with the first model, this model considered the between-subject
>variations, though the contrast would be set in the same way to investigate
>the difference between any two time points.
>
>So could you give me some suggests which one is more suitable for a
>longitudinal vbm analysis.
>
>Best,
>Liang
>
>
>2013/12/9 Christian Gaser <[log in to unmask]>
>
>> Dear Liang,
>>
>> the reason to skip the modulation step for longitudinal data was that the
>> effect of modulation on longitudinal data is rather subtle. Because the
>> deformations due to non-linear normalization are the same for all time
>> points the modulation only depends on the affine component that is driven
>> by different brain sizes. You could try to define a covariate for total
>> brain size (TIV) in your design to covariate out this potential effect.
>> However, I don't expect too much change of your results...
>>
>> Best,
>>
>> Christian
>>
>> On Thu, 5 Dec 2013 16:24:22 +0800, Liang Wang <[log in to unmask]> wrote:
>>
>> >Hi Christian and SPM folks,
>> >
>> >I am working on a longitudinal data using VBM8. Following the VBM-manual
>> >and changing the input data format to adapt to a longitudinal analysis, I
>> >sucessfully got the warped normalized gray matter images (wp1mr*). By
>> >searching this mail list, I saw that a long time ago Christian mentioned
>> >that it was not necessary to output the modulated GM volume (rather than
>> GM
>> >intensity) because the spatial normalization (= inverse modulation) is the
>> >same across multiple time points. It sounds reasonable. However, I still
>> >feel this measure (GM intensity, usually less than 1) was hard to be
>> >understood and the results were also hardly interpreted. I am wondering
>> how
>> >to make VBM8 to output the modulated images for the longitudinal analysis.
>> >I tried to change the parameters stored in cg_vbm8_defaults.m, but it did
>> >not work. Thanks.
>> >
>> >Best,
>> >Liang
>> >
>> >--
>> >Liang Wang, PhD
>> >Institute of Psychology
>> >Chinese Academy of Sciences
>> >Beijing 100101, China
>> >
>>
>>
>>
>
>
>--
>Liang Wang, PhD
>Institute of Psychology
>Chinese Academy of Sciences
>Beijing 100101, China
>
|