Bias correction in SPM99 was pretty terrible. There was a conceptual
problem in the way it was done, but this was fixed for SPM2 and SPM5.
The difference relates to working with the entropy of the image
intensity distribution, or the entropy of the distribution of the
log-transformed intensities. If working with original image
intensities, then a correction needs to be included
(http://mathworld.wolfram.com/ChangeofVariablesTheorem.html). SPM99 did
not include this, so there was a tendency to try to scale the image so
that its intensities ware as small as possible (to try to obtain an
intensity distribution with a single peak at zero). Although the model
didn't explicitly use histograms, it effectively worked with intensity
distributions that were parameterized with a mixture of Gaussians.
Similar principles applied.
Best regards,
-John
-----Original Message-----
From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]]
On Behalf Of Ged Ridgway
Sent: Thursday, January 18, 2007 12:47 PM
To: [log in to unmask]
Subject: Re: [SPM] inhomogeneity correction
Hi Manish,
There is a recent review of bias correction methods by Hou, but it
doesn't analyse performance. http://dx.doi.org/10.1155/IJBI/2006/49515
Arnold et al evaluated six algorithms, including N3 and spm99. N3
appears the best to me, and spm99 doesn't do too well, but spm5 is
probably greatly improved. http://dx.doi.org/10.1006/nimg.2001.0756
I'm not aware of a more recent comparison, including N3, FAST, and
SPM5 (which would be useful...). Anyone else?
Best,
Ged.
Manish Dalwani wrote:
> Dear SPM'ers,
>
> Does anyone know about a good inhomogeniety correction tool (for high
> field MRI) which could be applied for/to VBM studies?
>
> Thanks,
> Manish Dalwani
> PRA
> Dept. of Psychiatry
> UCHSC
>
>
------------------------------------------------------------------------
> TV dinner still cooling?
> Check out "Tonight's Picks"
> <http://us.rd.yahoo.com/evt=49979/*http://tv.yahoo.com/> on Yahoo! TV.
|