3D nearest neighbours and identifying clumps (clusters)

[pauliehagan]

Hello list,
We are new to imagej and over the last few weeks have been testing the graph plugin to identify clumps or clusters of points on a plane. The points are randomnly generated (by a variety of methods) x y coordinates plotted on a conventional x y scatter plot (this is the image we analyze, attached if my explanation is nonsense). We use the graph plugin after analyze particle to generate a list of the clusters and the points which make up each cluster. We can then compare this with statistical models which attempt to predict the number of clusters with 1, 2, 3, 4, ...............n membership.
The output produced by the graph plugin is perfect as we can easily process the log in excel to identify the number of 1's 2's 3's etc. Then compare this with theory.
So far so good, the 2D picture is working out a treat.
Next stage is the analysis in 3D. It has been suggested that a simple modification of the graph plugin will do the job.
We have our list of x.y and z coordinates from the randomn generators so at this point it is not strictly necessary to analyse an image.
So is anyone aware of a plugin which will do this? or how graph might be modified to perform the 3D analysis? I guess that a 3D image would have to be subdivided into thin slices and do the nearest neighbour on each slice then put the sliced up volume back together again?
All this will lead to a better understanding of how clumped nanoparticles constitute a greater health hazard than low airborne concentrations of single nanoparticles.

Many thanks
Paul
 
 

[Marcel]

Hello Paul,

if you have x, y, z coordinates or coordinates measured with ImageJ i think you can do this kind of analysis easily with R and the "spatstat" package which offers a huge amount of point pattern analysis methods.

http://www.spatstat.org

The "spatstat" package embeds methods to measure nearest neighbour distances in 2d and 3d
(function "nndist") and detect random, regular or clumbed patterns with e.g. the K-function (Ripley's K-function) in 2d or 3d (functions "Kest", "K3est").

This manual gives a short overview:

http://www.spatstat.org/spatstat/

[Thomas Boudier]

Hi Paul,

Maybe you can have a look to this page for how to manage 3D objects and
distances :

http://imagejdocu.tudor.lu/doku.php?id=plugin:stacks:3d_roi_manager:start

there is a simple macro on how to do basic clustering in 3D.

best,

Thomas


Le 21/07/2012 12:08, pauliehagan a écrit :

> Hello list,
> We are new to imagej and over the last few weeks have been testing the graph
> plugin to identify clumps or clusters of points on a plane. The points are
> randomnly generated (by a variety of methods) x y coordinates plotted on a
> conventional x y scatter plot (this is the image we analyze, attached if my
> explanation is nonsense). We use the graph plugin after analyze particle to
> generate a list of the clusters and the points which make up each cluster.
> We can then compare this with statistical models which attempt to predict
> the number of clusters with 1, 2, 3, 4, ...............n membership.
> The output produced by the graph plugin is perfect as we can easily process
> the log in excel to identify the number of 1's 2's 3's etc. Then compare
> this with theory.
> So far so good, the 2D picture is working out a treat.
> Next stage is the analysis in 3D. It has been suggested that a simple
> modification of the graph plugin will do the job.
> We have our list of x.y and z coordinates from the randomn generators so at
> this point it is not strictly necessary to analyse an image.
> So is anyone aware of a plugin which will do this? or how graph might be
> modified to perform the 3D analysis? I guess that a 3D image would have to
> be subdivided into thin slices and do the nearest neighbour on each slice
> then put the sliced up volume back together again?
> All this will lead to a better understanding of how clumped nanoparticles
> constitute a greater health hazard than low airborne concentrations of
> single nanoparticles.
>
> Many thanks
> Paul
>
>    http://imagej.1557.n6.nabble.com/file/n4999521/2220_black_circles.jpg
>
>
>
> --
> View this message in context: http://imagej.1557.n6.nabble.com/3D-nearest-neighbours-and-identifying-clumps-clusters-tp4999521.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
>

--
  /**********************************************************/
     Thomas Boudier, MCU Université Pierre et Marie Curie,
     Modélisation Cellulaire et Imagerie Biologique (EE1),
     IFR 83, Bat B 7ème étage, porte 723, Campus Jussieu.
     Tel : [phone]   Fax : [phone]
/*******************************************************/

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html