frequency histogram for diffusion analysis

[Adam Huttenlocker]

Dear all,

Does anyone know of a script or protocol for producing a histogram of the
shortest distances between particle centroids and, say, 100
randomly-generated particles in the intervening space? I am not trying to
do a nearest-neighbor analysis (which only gives the average distance
between each existing particle in an image and its closest neighbor).
Rather, I am trying to generate a profile (histogram) that represents the
average distances between random points in the interstitial space and their
closest particle centroid (e.g., used in studies of Krogh diffusion).

If anyone out there has experience with this method, I'd greatly appreciate
a nudge in the right direction. Every search I attempt on Google brings me
to nearest-distance plugins (which is similar, but not quite the same).

Thanks!

-Adam Huttenlocker

Adam K. Huttenlocker, PhD
Postdoctoral Research Fellow
Department of Biology, University of Utah
& Natural History Museum of Utah
Salt Lake City, UT 84112

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

[Thomas Boudier]

Dear Adam,

Maybe you can have a look to :

http://imagejdocu.tudor.lu/doku.php?id=plugin:analysis:spatial_statistics_2d_3d:start

Best,

Thomas


On 12/12/14 09:02, Adam Huttenlocker wrote:

> Dear all,
>
> Does anyone know of a script or protocol for producing a histogram of the
> shortest distances between particle centroids and, say, 100
> randomly-generated particles in the intervening space? I am not trying to
> do a nearest-neighbor analysis (which only gives the average distance
> between each existing particle in an image and its closest neighbor).
> Rather, I am trying to generate a profile (histogram) that represents the
> average distances between random points in the interstitial space and their
> closest particle centroid (e.g., used in studies of Krogh diffusion).
>
> If anyone out there has experience with this method, I'd greatly appreciate
> a nudge in the right direction. Every search I attempt on Google brings me
> to nearest-distance plugins (which is similar, but not quite the same).
>
> Thanks!
>
> -Adam Huttenlocker
>
> Adam K. Huttenlocker, PhD
> Postdoctoral Research Fellow
> Department of Biology, University of Utah
> & Natural History Museum of Utah
> Salt Lake City, UT 84112
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
   /***************************************************************/
      Thomas Boudier, Associate Professor, UPMC,
      Université Pierre et Marie Curie, Paris, France.
      BioInformatics Institute (BII)/IPAL, Singapore.
/**************************************************************/

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

[Gabriel Landini]

On Thursday 11 Dec 2014 18:02:51 Adam Huttenlocker wrote:
> Does anyone know of a script or protocol for producing a histogram of the
> shortest distances between particle centroids and, say, 100
> randomly-generated particles in the intervening space? I

I think you can get that from the distribution of edges of the Minimal
Spanning Tree graph. There was a plugin to compute the MST posted recently.

http://imagejdocu.tudor.lu/doku.php?id=plugin:morphology:euclidean_minimum_spanning_tree:start

Regards

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

[Jeremy Adler]

You need to consider edge effects - particles close to edges will have a different of distances to nearest and random neighbours, simply because we know nothing about off image neighbours. This can be avoided with a guard area around the images - for a nearest neighbour analysis rejecting particles whose distance to the image edge is less than the distance to their nearest neighbour, though even this seems to bias the particle selection a little as particles with distant nearest neighbours will be more likely to be rejected. So a guard area that avoids measurements from particles closer to the edge than the likely greatest interparticle distance is really needed.

 It is hard to see how you could produce an unbiased distance measurement between randomly selected second particles unless the image includes the whole area of interest- any ideas ?

Out of interest what are you ultimately trying to measure ?

A related question is what distance will you use. The straight line distance is the simplest choice but is wrong if you are measuring distances between say objects on the cell membrane when the shortest distance within the membrane may be more relevant - it is certainly a way to distort single particle tracking measurements and produce underestimates of diffusion.

-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Adam Huttenlocker
Sent: den 12 december 2014 02:03
To: [hidden email]
Subject: frequency histogram for diffusion analysis

Dear all,

Does anyone know of a script or protocol for producing a histogram of the shortest distances between particle centroids and, say, 100 randomly-generated particles in the intervening space? I am not trying to do a nearest-neighbor analysis (which only gives the average distance between each existing particle in an image and its closest neighbor).
Rather, I am trying to generate a profile (histogram) that represents the average distances between random points in the interstitial space and their closest particle centroid (e.g., used in studies of Krogh diffusion).

If anyone out there has experience with this method, I'd greatly appreciate a nudge in the right direction. Every search I attempt on Google brings me to nearest-distance plugins (which is similar, but not quite the same).

Thanks!

-Adam Huttenlocker

Adam K. Huttenlocker, PhD
Postdoctoral Research Fellow
Department of Biology, University of Utah & Natural History Museum of Utah Salt Lake City, UT 84112

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

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

[Alexandre Dufour]

Dear Adam (and others),

I was wondering why you actually needed a random set of points to compare your particles against. Is this to compare clustering vs random spatial organisation? If so, there is an analytical method that does away with the simulation step, using Ripley’s K function. We have a post-doc in our lab (Thibault Lagache, cc-ed here) who studies precisely these distributions, and he recently published a nice paper about this [1].
In particular, he shows that the edge effects can be analytically taken into account, such that you can use the entire field of view without bias (the equations get really long and ugly, but it works like a charm).

His method is also ready-to-use as an Icy plugin [2], so it’s anyway worth a try (it works for colocalization as well [3])

Hope this helps,
Alexandre

[1] http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0080914
[2] http://icy.bioimageanalysis.org/plugin/Spatial_Analysis
[3] http://icy.bioimageanalysis.org/plugin/Colocalization_Studio



On 12 Dec 2014, at 10:12, Jeremy Adler <[hidden email]> wrote:

> You need to consider edge effects - particles close to edges will have a different of distances to nearest and random neighbours, simply because we know nothing about off image neighbours. This can be avoided with a guard area around the images - for a nearest neighbour analysis rejecting particles whose distance to the image edge is less than the distance to their nearest neighbour, though even this seems to bias the particle selection a little as particles with distant nearest neighbours will be more likely to be rejected. So a guard area that avoids measurements from particles closer to the edge than the likely greatest interparticle distance is really needed.
>
> It is hard to see how you could produce an unbiased distance measurement between randomly selected second particles unless the image includes the whole area of interest- any ideas ?
>
> Out of interest what are you ultimately trying to measure ?
>
> A related question is what distance will you use. The straight line distance is the simplest choice but is wrong if you are measuring distances between say objects on the cell membrane when the shortest distance within the membrane may be more relevant - it is certainly a way to distort single particle tracking measurements and produce underestimates of diffusion.
>
> -----Original Message-----
> From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Adam Huttenlocker
> Sent: den 12 december 2014 02:03
> To: [hidden email]
> Subject: frequency histogram for diffusion analysis
>
> Dear all,
>
> Does anyone know of a script or protocol for producing a histogram of the shortest distances between particle centroids and, say, 100 randomly-generated particles in the intervening space? I am not trying to do a nearest-neighbor analysis (which only gives the average distance between each existing particle in an image and its closest neighbor).
> Rather, I am trying to generate a profile (histogram) that represents the average distances between random points in the interstitial space and their closest particle centroid (e.g., used in studies of Krogh diffusion).
>
> If anyone out there has experience with this method, I'd greatly appreciate a nudge in the right direction. Every search I attempt on Google brings me to nearest-distance plugins (which is similar, but not quite the same).
>
> Thanks!
>
> -Adam Huttenlocker
>
> Adam K. Huttenlocker, PhD
> Postdoctoral Research Fellow
> Department of Biology, University of Utah & Natural History Museum of Utah Salt Lake City, UT 84112
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

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