[Insight-users] How to compute "PCA shape model"?

Zachary Pincus zpincus at stanford.edu
Mon May 29 14:55:34 EDT 2006


> Hi  Zach :
>     Thank you for your reply.
>     I have read the code in itkImagePCAShapeModelEstimator.txx . It  
> seems that the mean image and principal component outputs are
> generated in this class,and the inputs of this class is  
> TrainingImages.

This is correct.

> So i wondered what the  "signed distance maps a set of
> training shapes " mean in your reply, and how can i input these  
> images.

Use one of the Signed Distance Map filters available in ITK (there  
are several, check out the Doxygen class list) to convert binary  
training images (spanning the space of possible shapes you wish to  
segment) into signed distance maps, where each pixel's value is the  
distance from that pixel to the nearest edge of the binary object  
(and the pixel's value is negative if it is inside the object). This  
then represents each shape as the zero-level set of an image.

>     In the quoted code, the shape model defined by  
> CorpusCallosumMeanShape.mha and the first three principal
> components CorpusCallosumMode0.mha, CorpusCallosumMode1.mha and  
> CorpusCallosumMode12.mha. How can i generate
> these mha files? Can i write the principal component outputs of  
> itkImagePCAShapeModelEstimator class in  mha files
> directly?

That is correct. Or you could use them directly without writing them  
to files.



>
> Thanks
> Sunli
>
> ----- Original Message -----
> From: "Zachary Pincus" <zpincus at stanford.edu>
> To: "sun" <kacking123 at 163.com>
> Cc: "itk" <insight-users at itk.org>
> Sent: Sunday, May 28, 2006 3:41 AM
> Subject: Re: [Insight-users] How to compute "PCA shape model"?
>
>
>
> This is a longstanding gap in the documentation.
>
> In brief, you will need to generate signed distance maps of a set of
> training shapes, and pass them to the PCAShapeModelEstimator class.
> You may then choose to divide the returned "shape mode" images (note
> that output zero from the model estimator is the mean image) by the
> reported standard deviations, as described in the example.
>
> Let me know if more detail is required.
>
> Zach Pincus
>
> Program in Biomedical Informatics and Department of Biochemistry
> Stanford University School of Medicine
>
>
> On May 27, 2006, at 12:06 AM, sun wrote:
>
>> Hi all:
>>      I am using the filter "Geodesic Active Contours Segmentation
>> with Shape Guidance" to
>> segment my image now.But the quoted  code just reads in a "PCA
>> shape model" that has been pre-computed
>> earlier. I want to know how to compute such a shape model in  ITK.
>> Who can help me?
>>                                       sun li
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> 你 不 想 试 试 今 夏 最 “酷” 的 邮 箱 吗 ?
>> 蕴 涵 中 华 传 统 文 化 于 世 界 一 流 科 技 之
>> 中,创 新 Ajax 技 术,126 “D 计 划”火 热 体 验
>> 中 !
>> _______________________________________________
>> Insight-users mailing list
>> Insight-users at itk.org
>> http://www.itk.org/mailman/listinfo/insight-users
>
> _______________________________________________
> 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