Dear Nina,
On Tue, 25 Nov 2014 23:14:18 +0000, Nina Teroganova <[log in to unmask]> wrote:
>Dear SPM community,
>
>After running the 'Estimate & write' module with the VBM8 toolbox and looking at the 'Sample Homogeneity' (module 3), I have noticed that there is a bit of dura remaining on the peripherally of most brains. Is this normal? This includes outliers but seem to occur when looking at the "non outliers"?
>And if not, how can I adjust the parameters in VBM8 to remove these artefacts? In what proportion of the presence of dura within the GM mask is tolerated, if tolerated?
You can try to change the clean up setting to "thorough cleanup" which might probably help. If not, however, the remaining dura parts will only lead to higher variation in these areas and will therefore lower the statistical power to detect any effects. Thus, the worst effect will be that you can't find any effects in these regions which is not too bad because usually the dura should be out of your interest. Due to spatial smoothing the remaining dura might also influence values in its neighbourhood. Thus you should take a closer look at those regions.
>
>Also following the sample homogeneity step, when checking for outliers, what kind of structural features classify a brain as an outlier? When checking the GM mask the only obvious issue I saw was a little bit of remaining dura on the periphery, as well as "brightness" in subcortical regions (i.e. caudate, thalamus). However, when checking scans which had high covariance there was also a bit of dura remaining. So I'm a bit confused why VBM has marked these brains as outliers (is it to do with contrast intensity rather than structural irregularities or something else?)
The covariance check estimates the (normalized) covariance between all scans across the whole image. Thus, spatially small deviations might influence the covariance only by a little amount. However, the tool finds out which scans deviate a lot from the whole sample and point to data with motions artefacts, anatomical deviations etc. It is really helpful for checking a large amount of data and gives you a hint about deviating data that should be checked manually. However, even if data are outside of 2 SD, this not automatically means that you have to remove these data due to the common outlier criteria. But, it's a strong hint that you should carefully check these data in more detail to decide whether you can use it or not. You can also use the tool to rate you data into different quality classes and you can try an alternative analysis without some deviating data from the worst ratings.
Best regards,
Christian
>
>Thank you for your help in advance.
>
>-Nina
|