time domain curve fitter

[Simon Roussel]

Hi,
ImageJ has got a curve fitter.
I am looking for a "time-domain" or "z axis" image fitter.
I would like to have for each pixel of a stack, a fit (actually
gamma-variate)  along the z axis. It could return a stack with the same z
values, but where the pixels values would be predicted from the fit, or
alternatively another stack with the coefficients of the fit.
I would use that to analyse MRI data following first-pass in the brain of a
contrast agent to derive cerebral blood volume and mean transit time.
Does anyone know a plugin, macro which already does that or could be adapted ?
Thanks in advance,
Cheers,
Simon

[Daniel Glen]

Outside of the ImageJ domain, software for FMRI/MRI analysis like  
AFNI or SPM can probably do what you want. I am in the group that  
develops AFNI. The particular program to do what you want is called  
3dDeconvolve  in the AFNI sofware. A gamma variate is one of the  
standard choices for fitting FMRI data over time with stimulus  
events. Other options include tent functions, stick functions and non-
linear solutions via 3dNLfim.  To do this over the z-direction  
instead should require a small trick to transform the z into time for  
this analysis. You can check http://afni.nimh.nih.gov/afni/doc/edu/ 
afni_3dDeconvolve for more information.

On Nov 14, 2006, at 10:14 AM, Simon Roussel wrote:

> Hi,
> ImageJ has got a curve fitter.
> I am looking for a "time-domain" or "z axis" image fitter.
> I would like to have for each pixel of a stack, a fit (actually
> gamma-variate)  along the z axis. It could return a stack with the  
> same z
> values, but where the pixels values would be predicted from the  
> fit, or
> alternatively another stack with the coefficients of the fit.
> I would use that to analyse MRI data following first-pass in the  
> brain of a
> contrast agent to derive cerebral blood volume and mean transit time.
> Does anyone know a plugin, macro which already does that or could  
> be adapted ?
> Thanks in advance,
> Cheers,
> Simon

[Barboriak, Daniel]

Hi Simon,

We are fairly close to releasing a plugin that does some but probably not
all that you want.  I'm assuming, perhaps incorrectly, that you are doing
T2-weighted dynamic susceptibility contrast MRI.  Our plugin called DSCoMAn
(Dynamic Susceptibility Contrast MR Analysis) will be able to convert image
intensity DSC MRI data to delta R2 images.  Because we are studying brain
tumors, we are analyzing the data using the Boxerman Weisskoff method.  For
our inital release, users will also have the option of making crude maps of
rCBV and rMTT using the raw delta R2 images.  

Here is the URL with a description:

http://dblab.duhs.duke.edu/modules/news/article.php?
com_mode=nest&com_order=1&storyid=14

This is (I hope) only a couple of weeks away.

ImageJ has it's own simplex gamma variate fitting routine (Analyze ...
Tools ... Curve Fitting).  This could be adapted for your purposes,
although to be honest I am persuaded by the Li's argument (see
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?
db=pubmed&cmd=Retrieve&dopt=AbstractPlus&list_uids=14684217&query_hl=14&itoo
l=pubmed_docsum ) that the formula would need to be recast or a local
density random walk distribution would be better to use.  This will likely
be a future direction for our software.

Thanks,

--db

Daniel Barboriak, MD
Neuroradiology
Duke University Medical Center

 
On Tue, 14 Nov 2006 10:14:02 -0500, Simon Roussel <[hidden email]>
wrote:

>Hi,
>ImageJ has got a curve fitter.
>I am looking for a "time-domain" or "z axis" image fitter.
>I would like to have for each pixel of a stack, a fit (actually
>gamma-variate)  along the z axis. It could return a stack with the same z
>values, but where the pixels values would be predicted from the fit, or
>alternatively another stack with the coefficients of the fit.
>I would use that to analyse MRI data following first-pass in the brain of a
>contrast agent to derive cerebral blood volume and mean transit time.
>Does anyone know a plugin, macro which already does that or could be

adapted ?
>Thanks in advance,
>Cheers,
>Simon
>========================================================================

[Simon Roussel]

Hi Daniel,
Thanks for your reply,
I am looking forward for your tool to analyse my T2*-DSC MRI.
I am actually also working on C6 brain tumors on rats with BBB alteration.
I will thus appreciate your implementation of the Boxerman Weisskoff method.
For stroke studies crude maps of CBV and MTT might be a start, before going
further to some modeling.
I'll have a look to the Li paper you mentionned.
I am also starting to work with DTI, so I'll have a look also to JDTI.
Thanks again, you have given me quite some perspectives.
Cheers,
Simon



On Wed, 15 Nov 2006 09:42:30 -0500, Daniel Barboriak <[hidden email]>
wrote:

>Hi Simon,
>
>We are fairly close to releasing a plugin that does some but probably not
>all that you want.  I'm assuming, perhaps incorrectly, that you are doing
>T2-weighted dynamic susceptibility contrast MRI.  Our plugin called DSCoMAn
>(Dynamic Susceptibility Contrast MR Analysis) will be able to convert image
>intensity DSC MRI data to delta R2 images.  Because we are studying brain
>tumors, we are analyzing the data using the Boxerman Weisskoff method.  For
>our inital release, users will also have the option of making crude maps of
>rCBV and rMTT using the raw delta R2 images.  
>
>Here is the URL with a description:
>
>http://dblab.duhs.duke.edu/modules/news/article.php?
>com_mode=nest&com_order=1&storyid=14
>
>This is (I hope) only a couple of weeks away.
>
>ImageJ has it's own simplex gamma variate fitting routine (Analyze ...
>Tools ... Curve Fitting).  This could be adapted for your purposes,
>although to be honest I am persuaded by the Li's argument (see
>http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?
>db=pubmed&cmd=Retrieve&dopt=AbstractPlus&list_uids=14684217&query_hl=14&itoo
>l=pubmed_docsum ) that the formula would need to be recast or a local
>density random walk distribution would be better to use.  This will likely
>be a future direction for our software.
>
>Thanks,
>
>--db
>
>Daniel Barboriak, MD
>Neuroradiology
>Duke University Medical Center
>
>  
>On Tue, 14 Nov 2006 10:14:02 -0500, Simon Roussel <[hidden email]>
>wrote:
>
>>Hi,
>>ImageJ has got a curve fitter.
>>I am looking for a "time-domain" or "z axis" image fitter.
>>I would like to have for each pixel of a stack, a fit (actually
>>gamma-variate)  along the z axis. It could return a stack with the same z
>>values, but where the pixels values would be predicted from the fit, or
>>alternatively another stack with the coefficients of the fit.
>>I would use that to analyse MRI data following first-pass in the brain of a
>>contrast agent to derive cerebral blood volume and mean transit time.
>>Does anyone know a plugin, macro which already does that or could be
>adapted ?
>>Thanks in advance,
>>Cheers,
>>Simon
>>========================================================================
>========================================================================