Jeff, Colin,
my 2 cents (as repeatedly voiced before, so just for the records): there
is a systematic difference between the datasets you want to compare,
which is a problem. We do not know how big it is (I agree it is not
likely going to make a big difference), but from a theoretical point of
view, the problem is that you cannot know how big the problem is as the
difference between acquisitions is inherently linked with the group
difference you are interested in. In other words and as I said before,
what do you reply to a reviewer who asks you to demonstrate that the
group difference you report is not due to the difference in acquisition
parameters? There are probably also ways to investigate the magnitude of
the problem, but it may be tricky. I would therefore be careful with
conclusion drawn from such a scenario, and potentially investigate both
solutions (use-as-is or drop-a-slice), and would probably agree about
slice timing and derivatives.
Cheers,
Marko
Jeff Browndyke wrote:
> Thanks, Colin. My plan was to account for the different datasets in the
> group model, but I didn't know if there was anything in particular I
> needed to do prior to generating the data for 1st level models between
> the datasets. Better safe than sorry. My worries are lessened, though
> I am now going to need to do some rework on my preproc scripts.
>
> Regards,
> Jeff
>
> On Feb 2, 2014, at 4:17 PM, Colin Hawco <[log in to unmask]
> <mailto:[log in to unmask]>> wrote:
>
>> Adding or removing slices for TA calculation isn't really appropriate
>> or going to solve the problems. As it stands you have slight different
>> TAs. if you drop a slice (or add), one scan has slightly inaccurate
>> TA, which is probably worse.
>>
>> There is no reason that I am aware of why this slice number difference
>> should make any real difference in your statistical analysis. In a
>> standard GLM analysis, TA only comes into play if you do slice time
>> correction, which is not always necessary, or even advised at this
>> point. If you include HRF plus derivative it accounts for some of the
>> differences in slice timing (although I am sure at least some SPM
>> users are not going to be in favor of that approach).
>>
>> The number of slices difference will be obliterated during
>> preprocessing if you run normalization, which will put all scans into
>> a common space and coordinate frame. If I was you, I don't think I
>> would worry about the extra slice, just make a note of it in the
>> methods of your paper.
>>
>> Colin
>>
>>
>> On 2 February 2014 14:42, Jeff Browndyke <[log in to unmask]
>> <mailto:[log in to unmask]>> wrote:
>>
>> Hello all,
>>
>> I'm seeking to run some simple fMRI analyses between two different
>> datasets which used almost identical parameters (e.g., TR, voxel
>> size, etc) with the exception of the total slice number. One set
>> has 23 slices per volume, while the other has 24 slices per
>> volume. Since the slice number is used in estimating the TA, I
>> was wondering what might be the best approach to tackle this
>> problem...if at all possible. Could one just drop the last slice
>> from each functional volume in the 24 slice dataset? And, if so,
>> how would this be accomplished? Would it be a matter of adding a
>> slice to the 23 set? Perplexed.
>>
>> Thanks for any pointers or help.
>>
>> Warm regards to all,
>> Jeff
>>
>>
>
--
____________________________________________________
PD Dr. med. Marko Wilke
Facharzt für Kinder- und Jugendmedizin
Leiter, Experimentelle Pädiatrische Neurobildgebung
Universitäts-Kinderklinik
Abt. III (Neuropädiatrie)
Marko Wilke, MD, PhD
Pediatrician
Head, Experimental Pediatric Neuroimaging
University Children's Hospital
Dept. III (Pediatric Neurology)
Hoppe-Seyler-Str. 1
D - 72076 Tübingen, Germany
Tel. +49 7071 29-83416
Fax +49 7071 29-5473
[log in to unmask]
http://www.medizin.uni-tuebingen.de/kinder/epn/
____________________________________________________
|