Hi everybody,
for quite a while we have been comparing the cortical thickness between groups of patients and I hope we did everything as we should. Now, we have been trying to look for the correlation between cortical thickness and age and somewhere we seem to make a mistake. I know that it's difficult for you to find this mistake as there are many possible sources. Nevertheless, I hope that somebody has an idea ;-) Let me describe what we did:
1. We have a group of subjects, for all of the subjects we went through the steps of segementation, inflating, sphering, morphing ...
2. We edited a simple mtx-file named PatBas-Alter1.mtx (this is not the complete group, we just started with a few subjects to check the method). This file has only two columns, column 1 is the name of the subject in the SUBJECTS_DIR, column 2 is the age of the subject:
Patient_2_Baseline 33 Patient_4_Baseline 42 Patient_5_Baseline 27 Patient_6_Baseline 40 Patient_7_Baseline 35 Patient_9_Baseline 48 Patient_10_Baseline 31 Patient_12_Baseline 29 Patient_13_Baseline 33 Patient_14_Baseline 37 Patient_15_Baseline 36 Patient_16_Baseline 41 Patient_17_Baseline 41 Patient_18_Baseline 26 Patient_21_Baseline 30 Patient_22_Baseline 37
3. We run mris_glm with a script or directly from the command line: (we are using version 1.18, 2003/06/30 20:40:38 greve)
mris_glm --design $mtxfile.mtx --surfmeas thickness --hemi lh --trgsubj average7 --nsmooth 50 --gcv 0 0 1 --sigt sigt_$outputname-lh.w paint
4. We render sigt_$outputname-lh.w on the inflated surface of average 7. The result is surprising. Everywhere in the brain the cortical thickness shows a high positive correlation (t-values between 9 and 12) with age. Obviuosly, that's nonsense but I haven't seen any error message.
5. To check what may have happened we edited the mtx-file according to the motto the last will be the first, high age is assigned to young subjects and viceversa.
6. The resulting t-maps look very similar. They are not identical but again the cortical thickness shows a high correlation with (now faked) age, again we get t-values between 9 and 12.
Any ideas what might have happened? I'm sure that we made something very stupid but what can that be?? Our mtx-file has only two columns. Can that be a problem??
Thanks in advance for all suggestions
Claus
----------------------------------------------------------------------- Claus Tempelmann email: Claus.Tempelmann@medizin.uni-magdeburg.de Clinic for Neurology II MR department Tel.: +49-(0)391-6117183 OvG University Magdeburg +49-(0)391-6117177 Leipziger Strasse 44 39120 Magdeburg Fax: +49-(0)391-6117178 Germany