Hi all,
I'm attempting to use FreeSurfer 5.1 (version block below) to process some older T1 scans collected on a 1.5T magnet with a native voxel size of 0.9375x0.9375x1.5mm. Overall, FreeSurfer seems to deal with these scans just fine (resampling to 1 mm isotropic, etc.), but I'm a little concerned about the results of the built-in image normalization.
The potential issue is illustrated in an image that I've posted to http://imgur.com/NiEVQ. The isolated plot to the right shows the intensity histogram that we've observed for more recent 3T T1 scans, with an obvious WM peak at 110 and a GM peak around 65. Meanwhile, each of the six plots to the left is the intensity histogram for one of our 1.5T scans. Although they are internally consistent (with the upper-leftmost example being something of an outlier), the GM peak is much closer to the WM peak, with a mode around 85.
I'm wondering whether this reduced GM/WM separation is likely to influence the tissue segmentation process, and if so how I might best remedy the situation. At a guess, mri_normalize or mri_ca_normalize seem like useful functions, but I haven't found the documentation for either to be of much help. Any assistance or advice would be greatly appreciated! Thanks,
Dave Warren
Hi Dave,
that probably just reflects the intrinsic contrast of your images. Depending on sequence parameters gray matter can vary w.r.t. white matter intensity. Lower contrast will usually make things worse, but of course it also depends on things like the noise level and the sharpness of your data.
cheers Bruce
On Mon, 12 Sep 2011, David Warren wrote:
Hi all,
I'm attempting to use FreeSurfer 5.1 (version block below) to process some older T1 scans collected on a 1.5T magnet with a native voxel size of 0.9375x0.9375x1.5mm. Overall, FreeSurfer seems to deal with these scans just fine (resampling to 1 mm isotropic, etc.), but I'm a little concerned about the results of the built-in image normalization.
The potential issue is illustrated in an image that I've posted to http://imgur.com/NiEVQ.%C2%A0 The isolated plot to the right shows the intensity histogram that we've observed for more recent 3T T1 scans, with an obvious WM peak at 110 and a GM peak around 65. Meanwhile, each of the six plots to the left is the intensity histogram for one of our 1.5T scans. Although they are internally consistent (with the upper-leftmost example being something of an outlier), the GM peak is much closer to the WM peak, with a mode around 85.
I'm wondering whether this reduced GM/WM separation is likely to influence the tissue segmentation process, and if so how I might best remedy the situation. At a guess, mri_normalize or mri_ca_normalize seem like useful functions, but I haven't found the documentation for either to be of much help. Any assistance or advice would be greatly appreciated! Thanks,
Dave Warren
Post-doctoral Fellow Neurology Department University of Iowa Hospitals and Clinics davideugenewarren@gmail.com
-------- freesurfer-Linux-centos4_x86_64-stable-pub-v5.1.0 -------- Setting up environment for FreeSurfer/FS-FAST (and FSL) $Id: FreeSurferEnv.csh,v 1.79.2.1 2011/03/29 22:06:48 greve Exp $ FREESURFER_HOME /usr/freesurfer FSFAST_HOME /usr/freesurfer/fsfast FSF_OUTPUT_FORMAT nii.gz SUBJECTS_DIR /usr/freesurfer/subjects MNI_DIR /usr/freesurfer/mni FSL_DIR /usr/fsl
for 3T scans, you could try a new flag added in v5.0.0, -nuintensitycor-3T (include at the end of the recon-all command) which modifies the nu_correct parameters to optimize for 3T scans, per this paper:
http://web.mysites.ntu.edu.sg/zvitali/publications/documents/N3_NI.pdf
n.
On Mon, 2011-09-12 at 18:43 -0400, Bruce Fischl wrote:
Hi Dave,
that probably just reflects the intrinsic contrast of your images. Depending on sequence parameters gray matter can vary w.r.t. white matter intensity. Lower contrast will usually make things worse, but of course it also depends on things like the noise level and the sharpness of your data.
cheers Bruce
On Mon, 12 Sep 2011, David Warren wrote:
Hi all,
I'm attempting to use FreeSurfer 5.1 (version block below) to process some older T1 scans collected on a 1.5Tmagnet with a native voxel size of 0.9375x0.9375x1.5mm. Overall, FreeSurfer seems to deal with these scans just fine (resampling to 1 mm isotropic, etc.), but I'm a little concerned about the results of the built-in image normalization.
The potential issue is illustrated in an image that I've posted to http://imgur.com/NiEVQ. The isolated plot tothe right shows the intensity histogram that we've observed for more recent 3T T1 scans, with an obvious WM peak at 110 and a GM peak around 65. Meanwhile, each of the six plots to the left is the intensity histogram for one of our 1.5T scans. Although they are internally consistent (with the upper-leftmost example being something of an outlier), the GM peak is much closer to the WM peak, with a mode around 85.
I'm wondering whether this reduced GM/WM separation is likely to influence the tissue segmentation process, and ifso how I might best remedy the situation. At a guess, mri_normalize or mri_ca_normalize seem like useful functions, but I haven't found the documentation for either to be of much help. Any assistance or advice would be greatly appreciated! Thanks,
Dave Warren
Post-doctoral Fellow Neurology Department University of Iowa Hospitals and Clinics davideugenewarren@gmail.com
-------- freesurfer-Linux-centos4_x86_64-stable-pub-v5.1.0 -------- Setting up environment for FreeSurfer/FS-FAST (and FSL) $Id: FreeSurferEnv.csh,v 1.79.2.1 2011/03/29 22:06:48 greve Exp $ FREESURFER_HOME /usr/freesurfer FSFAST_HOME /usr/freesurfer/fsfast FSF_OUTPUT_FORMAT nii.gz SUBJECTS_DIR /usr/freesurfer/subjects MNI_DIR /usr/freesurfer/mni FSL_DIR /usr/fsl
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Bruce Fischl -
David Warren -
Nick Schmansky