[Insight-users] Re: pixels To cm..

[Luis Ibanez]

Hi  Yasser,

Well, in your first email you said that you already segmented
a tumor using ConfidenceConnected, and then computed the number
of pixels inside the tumor by using the RelabelComponentImage
Filter.


The expression:

                 "areaInPixels"

in my previous email was making reference to the

         "Number of Pixels inside the tumor"




    Regards,


        Luis



--------------------------------------------

yasser salman wrote:

> Hi luis ..,
> about *areaInPixels* did u mean No. of pixels in ROI
> or what..,
> yasser
> --- Luis Ibanez <luis.ibanez at kitware.com> wrote:
> 
> 
>>Hi Yasser,
>>
>>
>>1) For converting an Area measured in Pixels^2 to
>>    an area in mm^2 units you must multiply by the
>>    product of the pixel spacing:
>>
>>
>>    ImageType::SpacingType spacing =
>>image->GetSpacing();
>>
>>    const double factor = spacing[0] * spacing[1];
>>
>>    double areaInMillimeters = areaInPixels *
>>factor;
>>
>>
>>2) If you run the program in 3D, and count the
>>    number of pixels inside a region, that will be
>>    a measure of volume. If you want to convert it
>>to
>>    mm^3 then you must do a similar operation to the
>>    one in (1):
>>
>>    ImageType::SpacingType spacing =
>>image->GetSpacing();
>>
>>    const double factor = spacing[0] * spacing[1] *
>>spacing[2];
>>
>>    double areaInMillimeters = areaInPixels *
>>factor;
>>
>>     Note that you can run ITK algorithms in 3D
>>natively.
>>     You *do not* need to do slice-by-slice
>>processing.
>>
>>
>>3) In both 2D and 3D, if your original segmentation
>>method
>>    give you any indication of partial volume in the
>>pixels
>>    in the boundary, you could use it in order to
>>reduce the
>>    uncertainty of the volume estimation. This is
>>the case
>>    of LevelSet, for example.
>>
>>
>>
>>
>>NOTE: Given that pixels represent a discretization
>>of the
>>       original object, the sum of pixels inside a
>>region
>>       is just an *estimation* of the region or
>>volume.
>>       The estimation is good if the region is
>>convex and
>>       large since the discretization errors only
>>occur in
>>       the boundaries of the region.  Conversely, if
>>you have
>>       a region with concave regions and thin
>>protrusions,
>>       the sum of pixels will give you a poor
>>estimation of
>>       the region's volume. The estimation may be
>>off by a 100%
>>       of the real value, when the protrusions are
>>about 1 pixel
>>       wide. If you want to make a formal estimation
>>of the volume,
>>       you probably want to consult some of the
>>papers on evaluation
>>       of geometrical features in discrete spaces.
>>
>>       Look, of example a the book
>>
>>          "Geometry of Digital Spaces" by Gabor T.
>>Herman,
>>
>>
>>
>>
>>Regards
>>
>>
>>
>>     Luis
>>
>>
>>
>>
>>--------------------------
>>yasser salman wrote:
>>
>>
>>>Hi All,
>>>i already segment a tumor region in 2D using
>>>ConfidenceConnectedImageFilter and count the no of
>>>pixels using RelabelComponentImageFilter , 
>>>how can i convert these no of pixels into cm2 or
>>
>>mm2
>>
>>>(the corresponding area in cm or mm).
>>>another question plz, if i run this program to a
>>>volume, is the no of pixels in 3D (volume) equal
>>
>>to 
>>
>>>the sum of all pixels in segmented region in 2D
>>>slices,
>>>
>>>yasser
>>>
>>>
>>>
>>>		
>>>_______________________________
>>>Do you Yahoo!?
>>>Declare Yourself - Register online to vote today!
>>>http://vote.yahoo.com
>>>
>>>
>>
>>
>>
>>
>>
> 
> 
> 
> 		
> _______________________________
> Do you Yahoo!?
> Declare Yourself - Register online to vote today!
> http://vote.yahoo.com
> 
>