[Insight-users] EPI Ground Truth

Arash Jahangir arash at vije.ca
Wed Dec 21 21:03:52 EST 2005


Hi Martin,
In a way you are right about this being off topic.  But I believe you are 
wrong about this not being of general interest.  A good chunk of  ITK is 
concerned with registration and many people use it for non-rigid 
registration.  Without ground truth, it is very hard to evaluate the 
performance of these algorithms or applications.  So I believe the topic is 
of interest to the list, even though it is not strictly an ITK subject 
during the design phase (it will be during the implementation phase).

Arash


----- Original Message ----- 
From: <kavec at messi.uku.fi>
To: "Arash Jahangir" <arash at vije.ca>
Cc: "Insight Users" <insight-users at itk.org>
Sent: Wednesday, December 21, 2005 4:26 PM
Subject: Re: [Insight-users] EPI Ground Truth


> Hi,
>
> you should consider first talking to an MRI physicist in order to confront 
> what
> you would like to do. It is possible that there are people on this list 
> who
> could provide you some insight, but I am afraid, it is quite off topic for
> majority.
>
> Martin
>
> Citát Arash Jahangir <arash at vije.ca>:
>
>> Hi all,
>> I'd like to generate ground truths for the purpose of evaluating brain 
>> EPI
>> registration algorithms.  I'd like your feedback and advice about this as
>> well as any resources on the net you may think are useful to the task.
>>
>> 1. One approach is to consider a T2 image as a good proxy for EPI.  The
>> challenge then is to create a realistic distortion that can serve as a
>> synthetic EPI.  Obviously the intensity values will not be quite the 
>> same.
>> So the question here is then:
>>    1a) Are the intensity variations between T2 and EPI close enough to 
>> make a
>> warped T2 a good synthetic EPI?
>>    1b) What is a good approach to create a realistic warp to transform T2 
>> to
>> EPI?
>>
>>
>> 2)  The second approach is to collect a set of registered EPI-T2 (or EPI 
>> to
>> any other undistorted MRI image) set that human experts believe are 
>> excellent
>> registrations.  Then generate a statistical model for the typical warps 
>> that
>> undistort the EPI images.  Finally use one or more corrected (i.e.
>> undistorted/registered) EPI images as ground truth and apply the 
>> synthetic
>> warp to it to get a distorted EPI.  The issues here are:
>>    2a) Is this a sound approach?  What normalization issues should I 
>> consider
>> before applying a generic warp field based on statistics on an 
>> undistorted
>> image to produce a synthetic distorted EPI?
>>    2b) Is there a database of registered EPI images?
>>    2c) Are the correction warp fields for those EPI images available and 
>> if
>> not, how can I recover it
>>
>> 3.  I guess I should have asked this first :)  what work has already been
>> done in this area?  Are there programs or databases the provide warped 
>> and
>> undistorted EPI ground truths?
>>
>>
>> Any and all feedback is highly appreciated.
>>
>> Arash
>
>
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users
> 



More information about the Insight-users mailing list