Hello Matt,

It seems you do not want to ignore information from the second scan of the 
patient group but I am not absolutely sure what the goal is.

If it is a cross-sectional group difference you want to focus I see a number 
of possibilities.

(A longitudinal analysis does indeed not make much sense if there is no 
control group you have monitored. Exception: your insular atrophy rate is 
extremely high, in this case some atrophy rates from published values on 
healthy subjects could serve as reference).

1. Calculate an average of your insula volume for each patient from both scans 
- you can to some scan-rescan-precision statistics as the two volume should 
not differ too much -  and proceed with the statistics with this average 
insular volume.

2. Do the group statistics twice (P1-CON and P2-CON).
3. Or just ignore the first scan. If acquired atrophy should be more distinct 
in the second scan.

Calculating the insular volume from both scans:

I do not see a real advantage in averaging the 2 images - as the normalized 
images will not be exactly the same you will rather get some vagueness in 
your edges/tissue boundaries. I am not sure if after segmentation this 
results in an 'average local volume'.

Calculating it from one scan:
If the normalization procedure is only affine and you take the total of voxels 
within your mask (which is in the same space) you will get a value that you 
could label 'insular volume'. This procedure however brings the indivduals' 
brains to a standard size so total GM or (unnormalized) total brain volume or 
total intracranial volume should be included into the statistics, or ratios 
be calculated.

If also non-affine normalization is used then some modulation step seems 
necessary to account for local volume changes.

Best regards

P. G. Saemann
Max-Planck-Institute of Psychiatry
80804 Munich
Phone: 0049-89-30622-413