Hi,
On 19 Aug 2007, at 15:14, Tamara Cristescu wrote:
> Hi,
> I would be very grateful if anyone would have a solution for my
> problems
> with FAST. I never used it before but I have to run some rat
> volumetric
> analyses.
> System: FLS 4.0 running on Fedora.
> I have high resolution rat T2 data acquired with a 3D RARE protocol
> on a 7T
> Brucker. We use a special surface coil made of four separate
> surface coils
> (parallel imaging system).
> Image resolution is quite high, 0.160 x 0.160 x 0.160mm, covering
> the rat
> hippocampus in 40 slices in Z plane. When we reconstruct from
> Brucker format
> to Analyze we rescale the image by a factor of 10 (otherwise the
> analysis
> software can’t cope with it). This seems quite a common trick based
> on some
> messages I saw on the FSL mailinglist. The final image is
> 1.60x1.60x1.60mm,
> matrix remains the same as the original but the FOV gets larger.
> Bet works
> very nice, using the default Bet parameters but I decrease the
> fractional
> intensity threshold to 0.1. If I use 0.5 Bet is too strict and I
> loose bits
> of the brain. At 0.1 I get some junk tissue around but it’s fairly
> limited.
> However, I experience some problems with FAST. First of all I don’t
> get a
> full output but only the total CSF (0.7 seems to be quite a lot of
> CSF).
Make sure you turn on the partial volume modelling with
-e -ov
and then use fslstats on each PVE output with
-m -v
and * the 1st and 3rd outputs together to get total volume of that
tissue.
> I
> need the total gray and white matter as well as the total brain
> volume.
> Also, the final segmented image doesn’t look right. Although I
> don’t have
> experience with FAST, I can see that the grey and white matter
> segmented
> tissues don’t respect the normal anatomical distributions.
> I pasted below the FAST output. Any diagnostic would be very
> helpful. Right
> now I have two hypotheses: failure with Bet (too much junk tissue
> around
> confuses FAST) or initial change of voxel size.
If the brain after your voxel resizing is ellipsoidal, consider
changing one of the voxel dimensions so that it appears more
spherical - that may help BET.
Cheers.
>
> Yours,
> Tamara
>
>
> =============================================================
> /usr/local/fsl/bin/fast -t2 -c 3 -ob -od
> /mnt/Lacie_Big_Disc/barrie/Tamara/w_test_imageJ/test01_33_2_brain
> /mnt/Lacie_Big_Disc/barrie/Tamara/w_test_imageJ/test01_33_2_brain
> 202967.5 111304.6 530429.3
>
> FAST - FMRIB's Automated Segmentation Tool Version 3.53
>
> Image: /tmp/fsl_z4VOfJ_test01_33_2_brain
>
> T2-weighted image
> Imagesize : 150 x 150 x 40
> Pixelsize : 1.6 x 1.6 x 1.6
>
> Initial K-means segmentation...
> 8 main iterations ...
>
> Segmentation done successfully!
>
> Calculation time 62 seconds!
>
> Write segmentation image /tmp/fsl_z4VOfJ.fastout/QQQQQ_seg.hdr
> Successfully!
> Write bias field image /tmp/fsl_z4VOfJ.fastout/QQQQQ_bias.hdr
> Successfully!
>
> Class: CSF tissue 1 tissue 2
> brain percentage
> Volumes: 0.759717
>
> Finished
------------------------------------------------------------------------
---
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
------------------------------------------------------------------------
---
|