Amar,
Our experience in this area would suggest that interictal hypoperfusion
in SPECT is very difficult to detect quantitatively. I would think
that 37 normals would be enough to give you some statistical
power in your database. So for starters I would say that you're not
seeing much because there's not much to see in the original data
itself.
One question I have about your study is what kind of statistical power
you actually have on the interictal patient side of things. Using single
patient scans your statistical power would be very low - perhaps this
plays a role in the 0.998 set level you're seeing. As an intermediate
step I would suggest trying to find a group of patients with concordant
surgical sites and combining them as a group, and comparing this
group to the set of normals using your method. If this experiment
identifies the seizure focus, then your problem is likely to be
related to statistical power and using single patient scans. If not,
then your analysis may be off.
The only other possible problem I can see is that because the
interictal hypoperfusion phenomenon is of such low amplitude it
is possible that oversmoothing the image could obscure it. Your
kernel may be appropriate, but a smaller one might be worth a
try.
Ben
------------------------------------------------------------------
Benjamin H. Brinkmann, Ph.D.
Dept. of Nuclear Medicine
Centerplace 830 Phone: (507)266-0798
Mayo Clinic and Foundation FAX: (507)284-9434
Rochester, MN 55905 [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|