[Insight-users] how to generate the mean shape and principal mode images for GeodesicActiveContourShapePriorLevelSetImageFilter

[Zachary Pincus]

Hello,

You can find an up-to-date list of all the classes in the ITK CVS at:
http://www.itk.org/Doxygen/html/classes.html

Specifically, the class for creating signed distance maps are:
ApproximateSignedDistanceMapImageFilter
and
SignedDanielssonDistanceMapImageFilter

You'll need a relatively recent CVS checkout to get these filters.

Also note that the input to these filters needs to be an image where  
the inside of the shape is all one intensity, and the outside is zero.  
The distance map filters won't work if the image is just of an outline  
with zero intensity pixels on both sides of the outline.

Finally, I do not believe that the "ActiveShapeModel" classes have  
anything to do with the  
GeodesicActiveContourShapePriorLevelSetImageFilter and related family  
of methods. The ActiveShapeModel classes implement an alternative  
(landmark-based, if I recall) way of incorporating shape models into  
image segmentations. (See the references given in the documentation of  
these classes). To generate shape models for the  
ShapePriorLevelSetImageFilter classes, you make a set of distance maps,  
then feed them into the ImagePCAShapeModelEstimator class.

Zach

On Jun 6, 2005, at 11:36 AM, <kingaza at gmail.com> wrote:

> Hi Zachary, all
>
> thx for your guide. and i can run it.
> but now i have still some questions.
>
> from DsnielssonSignedDistanceMap, the distances are 0 in the contour
> and more than 0 outside. it seems not SIGNED.
> btw, i can not find the ApproximateSignedDistanceMap, i don't know  
> where it is.
>
> I also try itkActiveShapeModelCalculatorTest, and i want to know:
> a) how can i get the mean image besides the meanshape
> b) what's the shape from
> ActiveShapeModelGradientSearchMethod::GetNewShape()? is it the shape
> in the input image?
>
> any tip is appreciated!
>
>
> On 6/5/05, Zachary Pincus <zpincus at stanford.edu> wrote:
>> Hello,
>>
>> You'll want to use the ImagePCAShapeModelEstimator class to get the
>> mean and principal modes of a set of signed distance maps.
>>
>> To generate signed distance maps, you feed (properly registered)  
>> binary
>> masks of your shapes of interest into one of the SignedDistanceMap
>> classes (either the ApproximateSignedDistanceMap class which is fast,
>> or the DsnielssonSignedDistanceMap class, which is exact.)
>>
>> I strongly suggest reading the Leventon paper where this method is
>> described in detail, as well:
>>
>> http://www.spl.harvard.edu:8000/pages/papers/leventon/cvpr00/ 
>> cvpr00.pdf
>>
>> Zach Pincus
>>
>> Department of Biochemistry and Program in Biomedical Informatics
>> Stanford University School of Medicine
>>
>>
>>
>> On Jun 4, 2005, at 10:56 AM, <kingaza at gmail.com> wrote:
>>
>>> Hi, all
>>>
>>> I find the asm/aam is a intersting method, but now i has some
>>> questions about the example.
>>> in the example, the mean shape and principal mode are files, but i  
>>> can
>>> not find how the files were generated. and i don't know how to use it
>>> in other occasion.
>>>
>>> please help me!
>>>
>>>
>>> --
>>> Regards,
>>> Kingaza
>>> _______________________________________________
>>> Insight-users mailing list
>>> Insight-users at itk.org
>>> http://www.itk.org/mailman/listinfo/insight-users
>>>
>>
>>
>
>
> -- 
> Regards,
> Kingaza
>