Dear BS,
The nonlinear registration that is currently in SPM only uses a small number
of parameters, so it is not able to estimate very detailed deformations. Dan
Rueckert's algorithm gives more detailed deformations, so it would be a
possibility. The HAMMER algorithm is one of the widely used approaches for
nonlinear registration
https://www.rad.upenn.edu/sbia/software/index.html#hammer , and most users
say it gives good results. In order to obtain the most meaningful
parameterisation of brain shapes, then I would opt for an algorithm such as
LDDMM - but I'm not aware of there being a freely available implementation of
this.
Among the next set of updates for SPM5, there will be some new nonlinear image
registration software that could be useful for your project. I'm currently
making a number of changes to the code, but I hope it should be available for
general use fairly soon. The algorithms are described in my recent
NeuroImage paper.
For the most unbiased results, I would suggest a template that is the average
shape and intensity of your subjects. Personally, I prefer to simultaneously
match GM with GM, WM with WM and background with background, rather than
match the intensities to some template image. The toolbox will use this
approach by default.
My favoured approach for analysing deformation fields involves the use of
multivariate pattern recognition procedures such as SVMs or Gaussian Process
models. Alternatively - depending on the aims of your analysis, you could
just run SPM on the Jacobian determinants - or do a VBM study that makes use
of the more accurate registration.
Best regards,
-John
On Wednesday 26 September 2007 10:24, you wrote:
> Respected Professor John Ashburner,
>
> My name is B S Mahanand, working as a visiting researcher in the
> Laboratory of Biomedical Engineering at Helsinki University of Technology,
> Espoo,FInland.
>
> I had previously mailed you regarding downloading of SPM5 and I could
> successfully download it with your help.I Thank you for your useful
> time and suggestion.
>
> Currently i am working on deformation field analysis of registered MR
> images.
>
> I would like to bring to your kind notice that,i am using Deformation
> based Morphometry method (DBM) proposed by you for my research.
>
> Personally I would like to sincerely thank you for publishing it and
> making it available
> to research community.
>
> Also i bring to your notice that currently i am using ADNI database
> for my research
>
> I have some queries regarding the usage of DBM
>
> 1.Sir I have a dilemma in the choice of atlas to which all subjects in
> the group are registered and i am thinking whether bias in the choice
> of atlas affects the results obtained using deformation-based
> morphometry.
> Kindly let me know whether to Use MNI152, Colin27, Alzheimer's Disease
> (AD) template from LONI,UCLA or any other atlas suggested by you
>
> 2. Secondly, regarding the choice of accurate non linear registration
> methods to use for analysis of deformation field using DBM.Kindly let
> me know whether to use SPM, AIR, Professor D.Rueckert free-form
> deformation toolkit or any other toolkit suggested by you.
>
> 3. Also i request you to let me know how to analyze efficiently the
> deformation fields produced after non linear registration. If you any
> source code to analyze deformation fields using DBM kindly give permition
> to use.
>
> In this regard, I kindly request you to give me suggestions and
> advices, so that i
> can proceed further with my research
>
> Thanking you,
>
>
> Sincerely,
>
> B.S Mahanand
> Visiting Researcher,
> HUT,
> Espoo,Finland
|