There is no calculation of variances or uncertainties in FAST, so you cannot really get anything except an estimate of the volume. I'm also not exactly sure what you can using these volumes for. If it is for a group comparison or a longitudinal study then the relevant factor would not be the variance/uncertainty/confidence-interval from a single image but would relate to the between-subject or within-subject changes, and these are best estimated by looking at the cross-subject or cross-timepoint variances. Your statisticians should be able to work out this information if you give them the FAST estimates.

On 14 Mar 2013, at 17:48, "Ho, Chang Yueh" <[log in to unmask]> wrote:

Hello,

I am using automated segmentation as described in your educational materials by skull stripping with BET and using FAST to segment white, gray and CSF volumes. We are doing a feasibility project in autistic children and have a small sample of children with autism and age matched children without. I understand there is an option for FSLstats to display the 2-98% interval for signal intensity values for a data set but is there any option for the measure of accuracy of the specific volume of the data set obtained by –m –v? For example a confidence interval of the segmented volume of gray matter? I believe this would help our statistician with calculations for significance and power.

Thanks for your help.

Chang