Hi Pouria
I don't really know anything about FSL, so I can't speak for that, but
for SPM the gray matter segments (c1*) are probability maps. In other
words, each voxel in a c1* images represents the likelihood that the
corresponding voxel in the original anatomical image in gray matter.
For a good segmentation, one would expect these values to be close to 1
in areas that are gray matter and close to 0 elsewhere. Given that most
of any given subject's head is not gray matter, it would follow that
there will be a lot of 0 value voxels in the c1* image. I think your
mistake is in assuming that a histogram of a c1* image only represents
values within the gray matter areas themselves, whereas in actuality it
represents values over the entire image.
-Neil
Siddharth Srivastava wrote:
> Can we also see the GM segmentations (3 views, as SPM shows it) for
> both cases that you illustrate below?
> sid.
>
>
> On Tue, Aug 25, 2009 at 12:17 PM, Mojabi, Pouria <[log in to unmask]
> <mailto:[log in to unmask]>> wrote:
>
> Hi again and thank you very much for your replay
>
>
>
> Please take a look at these:
>
>
>
> This is histogram of grey matter done, using SPM/EMS
>
>
>
>
>
>
>
>
>
> Same subject, using FSL 4.0 this is the grey matter histogram:
>
>
>
>
>
>
>
> You were right, the peak is at probability 1, not 0.5.
>
>
>
>
>
> In general, aren’t we expecting to see decay as we approach zero
> probability?
>
>
>
> Your thoughts are highly appreciated
>
>
>
> Thank you
>
>
>
>
>
> -P
>
>
>
>
>
> ------------------------------------------------------------------------
>
> *From:* SPM (Statistical Parametric Mapping)
> [mailto:[log in to unmask] <mailto:[log in to unmask]>] *On Behalf
> Of *Siddharth Srivastava
> *Sent:* Thursday, August 20, 2009 9:16 AM
> *To:* [log in to unmask] <mailto:[log in to unmask]>
>
> *Subject:* Re: [SPM] T1 image Segmentation
>
>
>
> Hi Pouria,
>
> The gray and white matter "T1 intensities" will be Gaussian. If
> you mask the T1
> with the classification maps , then you will see a bell shaped
> distribution for
> the gray and while matter intensities. The histogram for the
> classification would
> be as you described, since a good classification would have
> classification
> probabilities closer to 1 on the tissue class under consideration,
> and closer to zero
> otherwise. I cannot explain why you saw a peak at 0.5 though, in
> case you were
> looking at the histogram of the classification (and not of T1 .*
> xx_seg1 , say).
> sid.
>
> On Thu, Aug 20, 2009 at 7:39 AM, John Ashburner
> <[log in to unmask] <mailto:[log in to unmask]>> wrote:
>
> I'm not exactly sure what Koen included in his EMS model, but
> there may
> be a partial volume correction in there, which would change the
> model of
> the intensity distribution. Also, I don't really understand what you
> mean by p=0, p=1, p=0.5 etc. What is p?
>
> Best regards,
> -John
>
>
> On Wed, 2009-08-19 at 13:32 -0700, Mojabi, Pouria wrote:
> > Dear Experts,
> >
> > Hope all is well.
> >
> > I started developing this concern when comparing Segmentation results
> > (white matter, grey matter and CSF) between SPM (EMS) and FSL; When
> > looking at for instance Grey matter, as far as the histogram is
> concern,
> > I was expecting to see a Gaussian distribution, is this not a correct
> > assumption?
> >
> > I could see bell shape histogram for FSL segmented data, but I was
> > expecting to see the peak where probability is equal to 1 but the
> peak
> > was at p=0.5
> >
> >
> >
> > For SPM segmented data using EMS toolbox, I see a flat histogram with
> > minor peaks at p=1 and p=0, not sure why?
> >
> >
> >
> > I am entirely puzzled by this, based on my understanding of Gaussian
> > Mixture Models and Expectation maximization algorithm, you separate
> > three Gaussians (ideally speaking, eliminating bias field
> corruption),
> > then you build your probability mask based on the mean and standard
> > deviation of the calculated Gaussians, so histogram of the Segmented
> > data should somehow reflect this Gaussian nature, right ?
> >
> >
> >
> > Your thoughts on this are highly appreciated
> >
> >
> >
> > -P
> >
>
> --
> John Ashburner <[log in to unmask]
> <mailto:[log in to unmask]>>
>
>
>
>
|