David, Eric, etc.
> Since warping only the segmented gray matter still produced the same
> errors, would it be safe to assume that the skull stripped would do the same?
Yes it does, at least in my hands. It is a byproduct of the inhomogeneity
in the Siemens sequence. What happens is that since the voxel values are
reduced at the top of the brain less gray matter is extracted up top during
segmentation. Similarly with skull stripping the gray matter values are not
fixed and so do not normalize properly. Attached is the method I came up
with if you are interested.
Darren
Adjusted Normalization Steps to Correct for Signal Drop-off on T1 images
* Overview of Steps
* Change defaults (bias correction, segmentation, spatial
normalization)
* Bias Correction of original sf*.img file
* Create a Brain Mask in FSL
* Segment
* Normalize structural
* Adjust Defaults (The items listed below are required changes for
this method. Defaults not listed can be adjusted according to the
application.)
* Bias Correction
* Histogram bins = 1024
* Regularization = light
* Cutoff = 25 mm
* Segmentation
* Bias reg = medium (0.01)
* Cutoff = 25 mm
* Clean up partition = Don’t do cleanup
* Spatial Normalization (for . . . parameter estimation)
* weight template = (pick) default brain mask
* weight source = (pick) weight sources
* cutoff = 30 mm
* regularization = extremely heavy regularization (100)
* # of non-linear iterations : 16
* Spatial Normalization (for . . . writing normalized)
* voxel size: 1 x 1 x 1 (for structurals) 3 x 3 x 3 (for
functionals) [If you know that the 1x1x1 voxel size is selected, you
don’t need this step.]
* Way to wrap images: no wrap
* Sampling interval
* Must adjust at the Command Line in Matlab
* >>global defaults
* >>defaults.segment.estimate.samp = 2 [it was 3, need 2]
* Bias Correction (Use the button on the main window.)
* Select sf*.img
* This will create an msf*..img
* Create a Brain Mask in FSL
* Check that you are in the correct directory.
* At the prompt >> fsl
* A GUI comes up - select BET Brain Extraction
* BET Tool comes up – click on Advanced Options
* Select – generate binary brain img
* Unselect – non-brain matter removed
* Perform the extraction (click Apply). You may have to adjust the
Fractional Intensity Threshold, and the Threshold Gradient to get a
satisfactory extraction. This is left to the user.
* When done, click on FSL View – open the file you just
created and make sure it looks OK, i.e., covers most of the brain, no
major portions of brain missing, etc.
* Click Cancel
* Exit FSL
* Check Mask Orientation
* Return to SPM2
* Display the mask
* Make sure the mask is not flipped (if vox size: # is negative,
then it is flipped) (NOTE: This is true for out installation at
northwestern but may not be true for your setup of spm2.)
* If flipped, find resize(x) edit box in the Display window and put
a -1 in the edit box. Then click reorient image.
* Select the image to reorient, then OK
* Smooth the brain mask file using a 12mm kernel
* Look at the newly created sms^.img (brain mask)
* It should look fairly smooth.
* Segment
* Use the bias corrected brain generated in step 2
* Click Segment
* Choose msf*.img
* Already spatially normalized: No.
* Modality: T1 (or whatever)
* If you look at the images, you may now notice new gray signal at
the top of the brain and some skull visible.
* Normalize
* Click Normalize
* Click determine parameters and write normalized
* Template: Choose gray.mnc in the apriori folder.
* Source image is msf *.seg1.img
* Source weighting image is the smoothed brain mask
* Images to write – apply to sf *.img or msf * .img functional
images or the T1 structural image depending on the application.
VERY IMPORTANT –Always look at the created files to make sure everything
went OK!!
At 02:16 PM 8/22/2005, Kareken, David A. wrote:
>Hi Eric,
> Since warping only the segmented gray matter still produced the same
> errors, would it be safe to assume that the skull stripped would do the same?
>
> > -----Original Message-----
> > From: Eric Zarahn [mailto:[log in to unmask]]
> > Sent: Monday, August 22, 2005 2:04 PM
> > To: Kareken, David A.; [log in to unmask]
> > Subject: Re: [SPM] Spatial Normalization Problems w/ SPM2
> >
> >
> > Dear David,
> >
> > I have found that skull/scalp stripping along with providing
> > a little help to the starting estimates tends to eliminate
> > unreasonable warping due to spatial normalization in most
> > cases (normalizing to a skull/scalp stripped template).
> >
> > Best,
> > Eric
> >
> >
> >
> > ----- Original Message -----
> > From: "Kareken, David A." <[log in to unmask]>
> > To: <[log in to unmask]>
> > Sent: Saturday, August 20, 2005 10:34 AM
> > Subject: Re: [SPM] Spatial Normalization Problems w/ SPM2
> >
> >
> > > >I spent one weekend coming up with a complicated sequence
> > of steps to
> > > >minimize the distortions, which required adjusting the
> > defaults, bias
> > > >correction, skull stripping the brain, segmenting and then
> > > >normalizing. I'll be happy to send the steps to you if you
> > want, but
> > > >the bottom line
> > was
> > > >it didn't work all the time.
> > >
> > > Thanks, Darren. I might indeed be interested in the long route
> > > solution
> > you came up with.
> > >
> > > I did try doing just the EPIs, and was impressed how well
> > it worked.
> > > My
> > main concern here is the stretching of orbital signal. It
> > seems to want take preserved signal in the orbital area and
> > stretch it downward to fill the signal void. Maybe just
> > apply fewer non-linear itterations?
> > >
> > > By the way, our design required getting 4 MPRAGE scans on different
> > > days--
> > even after registering/averaging the four image volumes
> > (which markedly improved the inhommegeniety of which you
> > spoke to barely detectable visual levels), it still committed
> > the same sins.
> > >
> > > -----Original Message-----
> > > From: Darren Gitelman [mailto:[log in to unmask]]
> > > Sent: Fri 8/19/2005 7:11 PM
> > > To: Kareken, David A.
> > > Cc: [log in to unmask]
> > > Subject: Re: [SPM] Spatial Normalization Problems w/ SPM2
> > >
> > >
> > >
> > > Hi David:
> > >
> > > We have seen the exact same thing, kind of like high vacuum was
> > > applied to the top of the head. Actually worse than the
> > fact that this
> > > occurs, is
> > that
> > > it occurs differently for each subject so group
> > normalization sucks.
> > > This seems to be due to signal inhomogeneity from the coil, the
> > > sequence or whatever. We had the same problem with the
> > > Transmit/Receive and the 8-channel head coil. Our physicist
> > hasn't had
> > > much luck with adjusting the MPRAGE sequence per se (but see below)
> > > and we were unable to implement satisfactorily the MDEFT sequence
> > > written up by Ralf Deichmann. By the way even though the
> > MPRAGE images
> > > don't "look" that bad, if you do a bias correction on one
> > of the bad
> > > volumes and then plot the bias field the region requiring warping
> > > becomes obvious.
> > >
> > > Anyway I've tried a number of techniques to fix this. As you note
> > > segmenting the brain and normalizing the gray matter segment to the
> > apriori
> > > gray matter image can help, but doesn't always solve the problem
> > > because the gray matter segmentation is also affected by the
> > > inhomogeneity. So the gray matter at the top of the head is subtly
> > > deficient and then normalizes poorly.
> > >
> > > I spent one weekend coming up with a complicated sequence
> > of steps to
> > > minimize the distortions, which required adjusting the
> > defaults, bias
> > > correction, skull stripping the brain, segmenting and then
> > > normalizing. I'll be happy to send the steps to you if you
> > want, but
> > > the bottom line
> > was
> > > it didn't work all the time.
> > >
> > > The best solutions we have found are:
> > > 1) Normalize the Functional image and apply that to the T1. Yes the
> > > functional images have a geometric distortion relative to
> > the T1 but
> > > it isn't as bad as the artifact you get from normalizing
> > the T1. Yes
> > > the functional images have a much coarser voxel size so this is not
> > > the thing to do for VBM, but it works if the primary purpose is to
> > > normalize the functionals and produce moderately accurate
> > functional
> > > overlays. Also if you acquire fieldmaps you can minimize the
> > > functional volume distortions.
> > >
> > > 2) I should mention that we had originally acquired the
> > MPRAGE slices
> > > axially. It seems that acquiring it sagittally works better or at
> > > least moves the inhomogeneity away from the top of the
> > head. So this
> > > is what we are doing now for all T1's.
> > >
> > > We have also tried acquiring both T1 and T2 images for
> > multispectral
> > > segmentation and this sometimes improves the gray matter
> > segmentation
> > > and then the normalization.
> > >
> > > Anyway if you have further thoughts on this, or anyone else
> > does, or
> > > someone has an easily implemented T1 sequence that fixes
> > this problem
> > > on the Trio, I'd love to hear about it.
> > >
> > > Best regards,
> > > Darren
> > >
> > >
> > > At 05:55 PM 8/19/2005, Kareken, David A. wrote:
> > > >Hi all,
> > > >
> > > > I've been having a terrible time with spatially
> > normalizing some
> > > > T1 (MPRAGE from a Siemens Trio) images, and every single
> > subject has
> > > > the same problem. I've research this on the mail base,
> > and nothing
> > > > seems to help much.
> > > >
> > > > First, the whole brain images are stretched high at the top
> > > > (vertex), the occipital lobe/cerebellum are dragged down, and the
> > > > y-axis is abnormally compressed. The first and third problems
> > > > resolved when I segmented the brains, and registered gray
> > matter to
> > > > the gray apriori image, but the occipital lobe/cerebellum
> > is still
> > > > dragged down/distorted quite severely. That aspect seems
> > untouched.
> > > >
> > > > I seem to have the 10/04 code fix to spm_normalise.m that
> > > > addresses
> > the
> > > > Matlab 6.5 bug. Default masking doesn't seem to do
> > anything. Quite
> > > > frustrating!
> > > >
> > > > If you'd cc reply's to me, in addition to the mail base, I'd
> > appreciate
> > > > it. Thanks to any/all who can help.
> > > >
> > > >David K
> > > >
> > > >David A. Kareken, Ph.D., ABPP/ABCN
> > > >Associate Professor & Director of Neuropsychology
> > > >Department of Neurology (RI-1773)
> > > >Indiana University School of Medicine
> > > >Indianapolis, IN 46202
> > > >Tel: 317 274-7327
> > > >Fax: 317 274-1337
> > > >
> > > >
> > >
> > >
> > >
> > ----------------------------------------------------------------------
> > > ---
> > > Darren R. Gitelman, M.D.
> > > Cognitive Neurology and Alzheimer¹s Disease Center
> > > Northwestern Univ., 320 E. Superior St., Searle 11-470,
> > Chicago, IL 60611
> > > Voice: (312) 908-9023 Fax: (312) 908-8789
> > >
> > --------------------------------------------------------------
> > -----------
> > >
> > >
> > >
> > >
> >
> >
-------------------------------------------------------------------------
Darren R. Gitelman, M.D.
Cognitive Neurology and Alzheimer¹s Disease Center
Northwestern Univ., 320 E. Superior St., Searle 11-470, Chicago, IL 60611
Voice: (312) 908-9023 Fax: (312) 908-8789
-------------------------------------------------------------------------
|