Hi Jonathan,

On Mon, 14 Nov 2005, Jonathan Ipser wrote:

> I am part of a team that has been looking at structural changes in the brain
> following a clinical trial of typical antipsychotics for patients with
> schizophrenia. We are currently evaluating the suitability of using the fsl
> suite of brain imaging utilities to detect changes in caudate, lentiform and
> grey matter tissue volumes amongst responders to add-on treatment with
> ethyl-eicosapentaenoic acid (ethyl-EPA) for tardive dyskinesia. We have at
> our disposal dual-echo proton density and t2 MRI images of a total of 45
> patients (though for 17 we have only baseline images).
> 
> There are two major hypotheses which we would like to test:
> 
> (a) That the longer participants have been diagnosed with TD, the less
> caudate, lentiform and grey matter tissue they will possess, and
> 
> (b) That the neuro-protective effects of the medication resulted in
> relatively larger caudate, lentiform and grey matter tissue volumes in the
> medication than in the placebo groups. 
> 
> From what I can determine, the best way of proceeding to answer (a) is to
> run sienax on the baseline images for all participants, and to calculate the
> strength of the correlation of brain volume differences with duration of TD,
> while controlling for other possibly confounding variables (such as age and
> sex). Hypothesis (b) could be tested using siena on the pre and post
> medication brain images, stratifying participants by group. From what I
> understand, flow2tal would then be used to combine the images in each group
> for subsequent anaysis using the randomise tool (and I have yet to figure
> out how to construct the design matrix file for this analysis). 

Yes, you could use SIENAX or VBM to look at global, regional or local 
change in (eg) GM density; however it would be worth looking at the 
subcortical segmentation in the most recent version of FreeSurfer, which 
gives explicit segmentation of cubcortical structures and explicit 
estimation of cortical sufrace and thickness. Note that SIENA can only 
show where brain/non-brain boundaries move betweek two time points, so may 
not be useful on internal structures.

> I think I know how to calculate difference in grey matter volume for the two
> hypotheses. Sienax after all provides the amount of grey matter as default
> output, while for the longitudinal comparisons it is apparently possible to
> modify the source code for siena_diff, so that edge movement detection is
> limited to white matter (presumably one can then obtain an estimate of grey
> matter change through subtracting the white matter change from edge movement
> across the entire brain?). 

You could do this, yes, but the grey-white boundary for many of the 
internal structures is notoriously ill-conditioned because of poor g-w 
contrast and the lack (eg see thalamus) of a clear boundary.

> What is less clear to me is how to determine differences in the caudate and
> lentiform volumes for both hypotheses. What facilities are available within
> FSL to determine whether atropy (or growth) has taken place within these
> structures? Is is just a case of visualy inspecting the change image
> produced by siena, for instance, or are there other methods available?

See above - you might try FreeSurfer.

> Another concern I have is that siena seems to take excessively long in
> processing the input file, at round about 87 and 99 minutes for the two
> subjects I have tested it on. The article on using siena (S.M. Smith, Y.
> Zhang, M. Jenkinson, J. Chen, P.M. Matthews, A. Federico, and N. De Stefano.
> Accurate, robust and automated longitudinal and cross-sectional brain change
> analysis. NeuroImage, 17(1):479-489, 2002.) reported an average processing
> time of less than 1 hour. As this article was published approximately 3
> years ago, it was probably a lower spec machine than the 2.1 GHz AMD Athlone
> Linux machine (Debian 2.6.8) computer that I am using (with 512 Meg Ram and
> a 480 Meg Swap partition). I am also not trying anything too fancy, with the
> use of the following command line options: siena <image_1> <image_2> -t2 -f
> 0.4. Are there benchmarks available against which I can assess the
> performance of my computer (sienax takes approximately 25 minutes on this
> machine)?

Right - that seems reasonable, and is because we've improved SIENA a 
little since then, for example running the comparison forwards and 
backwards and taking the mean. Note that you might benefit from some more 
RAM, and you should definitely increase your swap space.

Hope this helps - cheers, Steve.

-- 
 Stephen M. Smith  DPhil
 Associate Director, FMRIB and Analysis Research Coordinator

 Oxford University Centre for Functional MRI of the Brain
 John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK
 +44 (0) 1865 222726  (fax 222717)

 [log in to unmask]  http://www.fmrib.ox.ac.uk/~steve