Hello,
in the siena-script the file '(A)_valid_mask_with_(B)' (and the counterpart
(B)_valid_mask_with_(B)' contains the information which flow-points (of all
points found) are analyzed.
In the example below a temporomesial_right.img/hdr (from Talairach) file is
taken into the halfway-position of the two images, some temporary files
temp_A and temp_B are used.
For a whole brain run, usually the siena_cal routine is called for both input
images and an average correction factor calculated. For now, the variable
corr is set to 1 WITHOUT calling the calibration routine at all, but maybe
this is not the most updated advice and the FSL programmers may comment.
Make sure the 'dotal' flag is set to 1 so the necessary ${A}_to_tal_inv.mat
and ${B}_to_tal_inv.mat are produced.
From here on only minor changes are necessary to make sure the result files
are named correctly, e. g. render-images, flow- and flowneg-images.
In the result render image you can see which points/regions were analyzed in
the end.
The method worked fine and produced reasonable results, however, the
calibration and validation remains difficult.
Interpretation: the good over-all power by of the whole brain analysis and the
robustness may suffer by a regional approach, this can also be seen from some
repositioning experiments. Also, the interpretation of the result as absolute
percentage brain volume change seems not allowed, however, for a relative
group comparison this worked well, and e. g. for Alzheimer's disease gave
sensible results.
The result does not represent (only) cortical atrophy (as e. g. the method of
Chen) but volume change of both grey and white matter as long as the
borderline to CSF is shifted.
many greetings,
Philipp
Max-Planck-Institute of Psychiatry
Munich
NMR RG
example:
# Temporomesial right
# ===================
# get and rename the template masks (in MNI) that are going to be flipped
into hw position
cp -f /usr/local/fsl/etc/standard/temporomesial_right.img temp.img
cp -f /usr/local/fsl/etc/standard/temporomesial_right.hdr temp.hdr
# get the mask into A (or B) position... (two runs of siena_diff, see below)
${FSLDIR}/bin/flirt -in temp -ref $A -out temp_A -applyxfm -init
${A}_to_tal_inv.mat
cp temp_A.img ${A}_valid_mask_with_${B}.img
cp temp_A.hdr ${A}_valid_mask_with_${B}.hdr
${FSLDIR}/bin/flirt -in temp -ref $B -out temp_B -applyxfm -init
${B}_to_tal_inv.mat
cp temp_B.img ${B}_valid_mask_with_${A}.img
cp temp_B.hdr ${B}_valid_mask_with_${A}.hdr
corr=1
From here on only minor changes are necessary to make sure the result files
are named correctly, e. g. render-images, flow- and flowneg-images.
In the result render image you can see which points/regions were analyzed in
the end.
The method worked fine and produced reasonable results, however, the
calibration and validation remains difficult.
Interpretation: the good over-all power by of the whole brain analysis and the
robustness may suffer by a regional approach, this can also be seen from some
repositioning experiments. Also, the interpretation of the result as absolute
percentage brain volume change seems not allowed, however, for a relative
group comparison this worked well, and e. g. for Alzheimer's disease gave
sensible results.
|