Measurement hierarchy

[karo03]

Hello ImageJ-experts and -gurus,
following the for me very helpful thread with Delaunay and Voronoi  
partitions I have now another problem difficult for me to solve:

See image http://picasaweb.google.com/lh/photo/CebTTCSWQsR5LjUzN9BRPg; 
I would like to measure several parameters like area, mean density,  
variation etc. for each region and for each small particle. As the  
region image is (in this case) an exo-skelteton derived from the  
particles, each region has a related particle.

Unluckily "Analyse particle" numbers each measurement different. To  
preserve the relationship I have to measure the regions and the  
particles both additionally with the label image to get a common label.

If I like to measure from the particles the area only I have divided  
the particle (binary) image by 255 and measured than the integrated  
density using the region rois. That was a solution for nerv fibres  
with axon.

Is there any idea to combine such measurement hierachy with less effort?

Best regards

Karsten Rodenacker
[hidden email]

[Gabriel Landini]

On Tuesday 02 December 2008 10:28:50 Karsten Rodenacker wrote:
> See image http://picasaweb.google.com/lh/photo/CebTTCSWQsR5LjUzN9BRPg;
> I would like to measure several parameters like area, mean density,
> variation etc. for each region and for each small particle. As the
> region image is (in this case) an exo-skelteton derived from the
> particles, each region has a related particle.
>
> Unluckily "Analyse particle" numbers each measurement different. To
> preserve the relationship I have to measure the regions and the
> particles both additionally with the label image to get a common label.

Hi Karsten,
You could try creating a results table that stores both, the start points of
the tiny particles and the starting points of the particle they belong to at
the same time (in different columns), so you can index the particles
according to the parent partice/partition.

To do this, you could: first extract the start points of the little particles
in the nucleus (binary image on the right), then one by one, reconstruct the
partitioned nucleus image (image on the top) based on the start points of the
small particles as seed. This will reconstruct a parent particle at a time
and now you can  extract its XStart YStart coordinates.

So now you have both the tiny particle start point and the parent particle
start point and can identify which one is inside which one.
Of course one has to store the tiny particles' starting points in an array as
the 2nd extraction might overwrite the table.

I hope it is clear.
Regards,

Gabriel

[Sami Badawi-2]

Hi Karsten,

Here is one solution using the the ShapeLogic plugin.
I assume that the image's color coding is using in inverted lookup
table, otherwise you have to change a few things.

Do an Image Calculation Or between you original image and the mask
that is in the top center of your image link.
Now you have the original image with the background being black and
the regions separated by thin black lines.

Call the ShapeLogic>> RGB Color Particle Analyzer
With these settings:
Max distance: 0
Min pixels: 0
Max pixels: 100000000
RGB R: 0
RGB G: 0
RGB B / Gray: 255
Background: checked
CountOnly: unchecked
ToMask: unchecked
DisplayInternalInfo: unchecked

Here is a link: http://www.shapelogic.org/particle.html

-Sami Badawi
http://www.shapelogic.org

On Tue, Dec 2, 2008 at 5:28 AM, Karsten Rodenacker <[hidden email]> wrote:

> Hello ImageJ-experts and -gurus,
> following the for me very helpful thread with Delaunay and Voronoi
> partitions I have now another problem difficult for me to solve:
>
> See image http://picasaweb.google.com/lh/photo/CebTTCSWQsR5LjUzN9BRPg; I
> would like to measure several parameters like area, mean density, variation
> etc. for each region and for each small particle. As the region image is (in
> this case) an exo-skelteton derived from the particles, each region has a
> related particle.
>
> Unluckily "Analyse particle" numbers each measurement different. To preserve
> the relationship I have to measure the regions and the particles both
> additionally with the label image to get a common label.
>
> If I like to measure from the particles the area only I have divided the
> particle (binary) image by 255 and measured than the integrated density
> using the region rois. That was a solution for nerv fibres with axon.
>
> Is there any idea to combine such measurement hierachy with less effort?
>
> Best regards
>
> Karsten Rodenacker
> [hidden email]
>

[karo03]

Thank you Gabriel,
I have still to learn result table handling and arrays!
However, your hints have helped me to avoid the second measure step  
with the labeled image. With the starting point I can extract the  
object number of both using the label image from your particle4/8  
routine and loo ... Of course region and particle intersect.

Karsten

Am 02.12.2008 um 12:16 schrieb Gabriel Landini:

> On Tuesday 02 December 2008 10:28:50 Karsten Rodenacker wrote:
>> See image http://picasaweb.google.com/lh/photo/ 
>> CebTTCSWQsR5LjUzN9BRPg;
>> I would like to measure several parameters like area, mean density,
>> variation etc. for each region and for each small particle. As the
>> region image is (in this case) an exo-skelteton derived from the
>> particles, each region has a related particle.
>>
>> Unluckily "Analyse particle" numbers each measurement different. To
>> preserve the relationship I have to measure the regions and the
>> particles both additionally with the label image to get a common  
>> label.
>
> Hi Karsten,
> You could try creating a results table that stores both, the start  
> points of
> the tiny particles and the starting points of the particle they  
> belong to at
> the same time (in different columns), so you can index the particles
> according to the parent partice/partition.
>
> To do this, you could: first extract the start points of the little  
> particles
> in the nucleus (binary image on the right), then one by one,  
> reconstruct the
> partitioned nucleus image (image on the top) based on the start  
> points of the
> small particles as seed. This will reconstruct a parent particle at  
> a time
> and now you can  extract its XStart YStart coordinates.
>
> So now you have both the tiny particle start point and the parent  
> particle
> start point and can identify which one is inside which one.
> Of course one has to store the tiny particles' starting points in an  
> array as
> the 2nd extraction might overwrite the table.
>
> I hope it is clear.
> Regards,
>
> Gabriel