Dear Gilberto,
FIRST can only be used on T1-weighted images as the intensity
model that it uses is based on training data that is all T1-weighted.
Hence I recommend you use a T1-weighted image to get your
segmentation, and if you then need it on the FLAIR image, do a
registration between the T1-weighted and FLAIR images (with 6 DOF)
and then transform the segmentation output into the space of the
FLAIR image.
All the best,
Mark
On 4 May 2011, at 20:35, Gilberto Sousa Alves wrote:
> Dear FSL experts,
>
> I am trying to run FIRST to segment brain anatomical structures with Flair images (T2 weighted volumes). My final goal is to get the segmentation for hippocampus and then to estimate right and left hippocampus volume.
> My parameters are as following:
> 1.5T GE Signa Horizon machine (General Electric, Milwaukee, WI, USA), TR/TE=10,000/89.1 msec; Matrix=128×128, FOV=30×24 mm, NEX=1, b=1000 sec/mm2; slice thickness=5 mm, number of slices=20 without gap;
>
> I have run the steps below according to the instructions of FSL website and FSL forum (“FIRST and hippocampus”, 27/10/2009). Images were cropped to remove the neck;
> Additionally, output images (named “headaxial”) were reoriented to axial in order to facilitate the registration:
>
> fslroi axial_t2.nii cropimage 112 282 67 332 0 19
>
> fslswapdim cropimage.nii.gz RL PA IS headaxial
>
> run_first_all -i headaxial.nii.gz -o hippocamp_seg
>
>
> The resulting output image after “run_first_all” was only one:
>
> "Hippocamp_seg_to_std_sub.nii.gz"
>
> The report of log files describe “no interior voxels to estimate mode” and also errors to find header file for “headaxial-L_Pall_corr”. The same error description is given to “headaxial-L_Pall_first”…
>
> I wonder if these problems are related to the quality of my Flair images. The slice is thickness is too large (5 mm) and because of that both coronal and sagittal slices look a little blurred.
>
> My final question is: Is FLAIR an appropriate image to get hippocampal volumes with FIRST? If not, what would be the alternative?
>
> Any suggestion will be of great help.
>
> Thanks is advance,
>
>
> Gilberto Alves
> PhD student, guest researcher,
> Laboratory of Neuroscience
> Wolfgang Goethe University- Frankfurt
>
|