Spatial relationship between two events in cells

[Krzysztof Berniak]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Dear List Members,

We are looking for an image analysis/cytometry tool that can perform
the following job:

We have a 3D image stack that consists of several hundred green and
red foci (they represent two nuclear processes that occur in many
little spots). We need to find the center of mass of each red 3D
focus, and measure the distance (in 3D) to the nearest green spot, and
repeat the process for all red spots in one 3D stack. Next, we need to
repeat this job in a large number of nuclei. Eventually, we need to
obtain a histogram of distances between the red spots and their the
nearest green spot neighbor in a large population of 3D objects.

Does anyone know a ready-to-use image analysis tool that can do this
job for us? We have started to work on our own algorithm, but maybe
there are solutions to this problem available already. Any advice
about a software package that could this work would be much
appreciated.

Thank you very much for your help,

Krzysztof Berniak, Jurek Dobrucki
Jagiellonian University

[Julio Vazquez]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi Krzysztof,

I believe Volocity (Improvision/PerkinElmer) should be able to do that. You may check their tutorial here:

http://www.perkinelmer.com/pages/020/cellularimaging/training/MeasuringDistancesVolocity.xhtml

Julio.

==
Julio Vazquez
Fred Hutchinson Cancer Research Center
Seattle, WA 98109-1024

http://www.fhcrc.org


----- Original Message -----
From: "Krzysztof Berniak" <[hidden email]>
To: [hidden email]
Sent: Monday, July 23, 2012 12:15:53 PM
Subject: Spatial relationship between two events in cells

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Dear List Members,

We are looking for an image analysis/cytometry tool that can perform
the following job:

We have a 3D image stack that consists of several hundred green and
red foci (they represent two nuclear processes that occur in many
little spots). We need to find the center of mass of each red 3D
focus, and measure the distance (in 3D) to the nearest green spot, and
repeat the process for all red spots in one 3D stack. Next, we need to
repeat this job in a large number of nuclei. Eventually, we need to
obtain a histogram of distances between the red spots and their the
nearest green spot neighbor in a large population of 3D objects.

Does anyone know a ready-to-use image analysis tool that can do this
job for us? We have started to work on our own algorithm, but maybe
there are solutions to this problem available already. Any advice
about a software package that could this work would be much
appreciated.

Thank you very much for your help,

Krzysztof Berniak, Jurek Dobrucki
Jagiellonian University

[Bonnaud, Fabien]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Dear Krzysztof, Julio,

Volocity is indeed perfectly suited for that sort of appplications.

It offers easy to use 3D image segmentation, compartmentalization and automatic 3D distance measurement, from centroid or edge of objects in any combination.
The distances are normalized to the size of each nucleus for more accurate measurements.
Each object's morphology can also be completely described.

The protocols set-up in Volocity can be applied to multiple images, as we use 3D automatic thresholding based on an Otsu method.

Our protocols can be ran on a batch of data without any scripting.
The results are exportable to Excel but analysis and histograms can also be obtained in Volocity.

Thanks for your question, should you require any additional information, please do not hesitate to contact me.

[hidden email]

All the best,

Fabien Bonnaud
Volocity Product Line Leader
PerkinElmer

Fabien Bonnaud

----- Original Message -----
From: Vazquez Lopez, Julio [mailto:[hidden email]]
Sent: Tuesday, July 24, 2012 05:29 AM
To: [hidden email] <[hidden email]>
Subject: Re: Spatial relationship between two events in cells

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi Krzysztof,

I believe Volocity (Improvision/PerkinElmer) should be able to do that. You may check their tutorial here:

http://www.perkinelmer.com/pages/020/cellularimaging/training/MeasuringDistancesVolocity.xhtml

Julio.

==
Julio Vazquez
Fred Hutchinson Cancer Research Center
Seattle, WA 98109-1024

http://www.fhcrc.org


----- Original Message -----
From: "Krzysztof Berniak" <[hidden email]>
To: [hidden email]
Sent: Monday, July 23, 2012 12:15:53 PM
Subject: Spatial relationship between two events in cells

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Dear List Members,

We are looking for an image analysis/cytometry tool that can perform
the following job:

We have a 3D image stack that consists of several hundred green and
red foci (they represent two nuclear processes that occur in many
little spots). We need to find the center of mass of each red 3D
focus, and measure the distance (in 3D) to the nearest green spot, and
repeat the process for all red spots in one 3D stack. Next, we need to
repeat this job in a large number of nuclei. Eventually, we need to
obtain a histogram of distances between the red spots and their the
nearest green spot neighbor in a large population of 3D objects.

Does anyone know a ready-to-use image analysis tool that can do this
job for us? We have started to work on our own algorithm, but maybe
there are solutions to this problem available already. Any advice
about a software package that could this work would be much
appreciated.

Thank you very much for your help,

Krzysztof Berniak, Jurek Dobrucki
Jagiellonian University

[Julio Vazquez]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Krzysztof, just an additional thought: I am not sure if Volocity will give you the individual distance values for each spot, which is what you would need to plot your Histograms, but probably only the averaged values for the whole population; for instance, you can get the average minimum distance of each green spot to the red spots (i.e. the average value of the distance of all green spots to their nearest red spot), or the average value of the distance of each green spot to all red spots, etc. Maybe other Volocity users, or a Volocity support person can comment on this. What Volocity (and other 3-D image analysis software such as Imaris) will give you, is the x,y,z location of the centroids of all the spots found. With those values, it should be possible and relatively easy for someone with programming skills to write a program to extract the center to center distances.  

ImageJ has a plugin to segment and measure 3-D objects (3D object counter). This will also list the x,y,z coordinateds of the geometric and/or intensity center, which you coudl use to make your own distance measurements:

http://rsbweb.nih.gov/ij/plugins/track/objects.html

I have never tried them, but FIJI may have additional tools for 3-D segmentation, and like ImageJ is free. A list of other software for 3-D analysis can be found here:

http://www.andrewnoske.com/wiki/index.php?title=Tomography_software

some of these (and maybe other) packages have been discussed on the confocal listserver.


--
Julio Vazquez,
Fred Hutchinson Cancer Research Center
Seattle, WA 98109-1024


http://www.fhcrc.org

===


On Jul 23, 2012, at 12:15 PM, Krzysztof Berniak wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Dear List Members,
>
> We are looking for an image analysis/cytometry tool that can perform
> the following job:
>
> We have a 3D image stack that consists of several hundred green and
> red foci (they represent two nuclear processes that occur in many
> little spots). We need to find the center of mass of each red 3D
> focus, and measure the distance (in 3D) to the nearest green spot, and
> repeat the process for all red spots in one 3D stack. Next, we need to
> repeat this job in a large number of nuclei. Eventually, we need to
> obtain a histogram of distances between the red spots and their the
> nearest green spot neighbor in a large population of 3D objects.
>
> Does anyone know a ready-to-use image analysis tool that can do this
> job for us? We have started to work on our own algorithm, but maybe
> there are solutions to this problem available already. Any advice
> about a software package that could this work would be much
> appreciated.
>
> Thank you very much for your help,
>
> Krzysztof Berniak, Jurek Dobrucki
> Jagiellonian University

[David Baddeley]

I'll break the trend and advocate that you continue to wo
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi Krzystof,

I'll break the trend and advocate that you continue to work on your own algorithm, rather than attempting to use pre-canned software such as Velocity or ImageJ. My reasoning is that without having developed and tested an algorithm yourself it's very difficult to understand it's limitations, and that the application you propose is likely to push the limits of any pre-canned algorithm. Several hundred foci sounds like you might be looking at replication foci, splicing speckles, or something of that nature. Several hundred foci also implies that you probably want to determine their position with sub-resolution accuracy (your ~800 nm z resolution is already ~10-20% of the size of the nucleus). I generally wouldn't trust a pre-canned algorithm to give you sub-resolution accuracy (it's definitely possible with careful use, but arguably requires an understanding of the underlying algorithm which would necessitate more time spent reverse engineering than you
 would have spent writing your own algorithm from scratch).

Generic point finding algorithms such as those found in ImageJ work by applying a global threshold and then calculating the geometric (or if you are lucky, intensity centre) of the objects above threshold. Assuming they calculate the intensity centre, they will usually do this without first subtracting the threshold. There are three major problems with this approach when measuring nuclear foci: 

- It is likely to be difficult if not impossible to find a single threshold which effectively separates all your objects, whilst still including enough voxels in every object for an accurate centroid estimation (you will probably want a ~5x5x5 voxel ROI to be above threshold for each spot if the centroids are to achieve sub-resolution accuracy)
- Even if you do manage this, centroid estimators show considerable bias. This bias can be substantially reduced by subtracting the threshold from the data prior to calculating the centroid, but this is not common practice. A pre-canned centroid estimator is thus unlikely to give you much better than ~ 0.5-1 voxel accuracy.
- Chromatic shifts are likely to be important on the distance scales you want to measure, and this is difficult to take into account in most existing packages.

The approach I have used in the past to identify and measure the positions of replication foci  and/or gene loci in the nucleus was to:
- use an iterative approach with a sinking threshold and removal of found points to determine the approximate positions of each locus. This is similar to the 'clean' algorithm used in astronomy to find stars and in PALM to identify single molecules, and enabled nearby foci and foci of varying brightness to be independently identified.
- fit a 3D Gaussian to each identified loci. This worked well for replication foci as they are typically sub-resolution, and gave comparable results to threshold-subtracted centroid estimation for gene loci even if they were just above the resolution limit. It is also surprisingly resilient against nearby loci, which might be expected to bias the fit.
- correct for chromatic shift using a separate measurement made using tetra-speck beads. In most cases it is sufficient to assume a constant shift, and ignore sample effects - this will typically give you an accuracy of 50-100nm. If you want to go below this, or if you have a particularly badly aligned microscope or a very high NA objective (our 1.49NA TIRF objective shows > 200 nm colour shifts towards the edges of the field), you will need to perform a spatially varying shift correction.

best wishes,
David


________________________________
 From: Krzysztof Berniak <[hidden email]>
To: [hidden email]
Sent: Tuesday, 24 July 2012 7:15 AM
Subject: Spatial relationship between two events in cells
 
*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Dear List Members,

We are looking for an image analysis/cytometry tool that can perform
the following job:

We have a 3D image stack that consists of several hundred green and
red foci (they represent two nuclear processes that occur in many
little spots). We need to find the center of mass of each red 3D
focus, and measure the distance (in 3D) to the nearest green spot, and
repeat the process for all red spots in one 3D stack. Next, we need to
repeat this job in a large number of nuclei. Eventually, we need to
obtain a histogram of distances between the red spots and their the
nearest green spot neighbor in a large population of 3D objects.

Does anyone know a ready-to-use image analysis tool that can do this
job for us? We have started to work on our own algorithm, but maybe
there are solutions to this problem available already. Any advice
about a software package that could this work would be much
appreciated.

Thank you very much for your help,

Krzysztof Berniak, Jurek Dobrucki
Jagiellonian University

[Barlow, Andrew]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Hi Krzysztof,
Just to clarify, Volocity will generate the exact output that you require, the individual distance from each red spot to the nearest green spot.  These distances can be displayed as feedback on the image, and plotted as a histogram. The distance can be calculated from either centroids, or nearest edges.  Distances can be calculated in voxels, calibrated units, or normalized to the longest axis of the containing nucleus.

We're really delighted with the developments we've made to Volocity recently and would be happy to demonstrate them so please contact us off list for further details.
Regards,
Andrew

> -----Original Message-----
> From: Confocal Microscopy List
> [mailto:[hidden email]] On Behalf Of Julio Vazquez
> Sent: 24 July 2012 18:40
> To: [hidden email]
> Subject: Re: Spatial relationship between two events in cells
>
> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Krzysztof, just an additional thought: I am not sure if Volocity will
> give you the individual distance values for each spot, which is what
> you would need to plot your Histograms, but probably only the averaged
> values for the whole population; for instance, you can get the average
> minimum distance of each green spot to the red spots (i.e. the average
> value of the distance of all green spots to their nearest red spot), or
> the average value of the distance of each green spot to all red spots,
> etc. Maybe other Volocity users, or a Volocity support person can
> comment on this. What Volocity (and other 3-D image analysis software
> such as Imaris) will give you, is the x,y,z location of the centroids
> of all the spots found. With those values, it should be possible and
> relatively easy for someone with programming skills to write a program
> to extract the center to center distances.
>
> ImageJ has a plugin to segment and measure 3-D objects (3D object
> counter). This will also list the x,y,z coordinateds of the geometric
> and/or intensity center, which you coudl use to make your own distance
> measurements:
>
> http://rsbweb.nih.gov/ij/plugins/track/objects.html
>
> I have never tried them, but FIJI may have additional tools for 3-D
> segmentation, and like ImageJ is free. A list of other software for 3-D
> analysis can be found here:
>
> http://www.andrewnoske.com/wiki/index.php?title=Tomography_software
>
> some of these (and maybe other) packages have been discussed on the
> confocal listserver.
>
>
> --
> Julio Vazquez,
> Fred Hutchinson Cancer Research Center
> Seattle, WA 98109-1024
>
>
> http://www.fhcrc.org
>
> ===
>
>
> On Jul 23, 2012, at 12:15 PM, Krzysztof Berniak wrote:
>
> > *****
> > To join, leave or search the confocal microscopy listserv, go to:
> > http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> > *****
> >
> > Dear List Members,
> >
> > We are looking for an image analysis/cytometry tool that can perform
> > the following job:
> >
> > We have a 3D image stack that consists of several hundred green and
> > red foci (they represent two nuclear processes that occur in many
> > little spots). We need to find the center of mass of each red 3D
> > focus, and measure the distance (in 3D) to the nearest green spot,
> and
> > repeat the process for all red spots in one 3D stack. Next, we need
> to
> > repeat this job in a large number of nuclei. Eventually, we need to
> > obtain a histogram of distances between the red spots and their the
> > nearest green spot neighbor in a large population of 3D objects.
> >
> > Does anyone know a ready-to-use image analysis tool that can do this
> > job for us? We have started to work on our own algorithm, but maybe
> > there are solutions to this problem available already. Any advice
> > about a software package that could this work would be much
> > appreciated.
> >
> > Thank you very much for your help,
> >
> > Krzysztof Berniak, Jurek Dobrucki
> > Jagiellonian University

[Daniel Sevilla]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
*****

Dear Krzysztof and Jurek,

The Huygens software may be very useful for you. Especially if you want
to analyze many images. The Huygens Object Analyzer is an extensive tool
which measures distances within channels as well as between channels.
Distances from CM to CM (center of mass), from CM to surface and/or
surface to surface are reported in real units by taking the sampling
into account. Nearest neighbor objects are automatically detected, if
requested. By analyzing all objects at once a direct histogram can be
obtained from the distances. Of course all statistics can be exported to
make your own histograms or graphs.

The Huygens software is very suitable for analyzing many images.
Template support across the Huygens tools allow you to carry out the
exact same analysis on multiple images. Additionally, scripting
functionalities allow more advanced users and programmers to make the
analysis fully automatic.

The best way to determine if the Huygens Software works for you is to
simply test the software, which you can download at www.svi.nl/download.
We will gladly provide you with a test license and manuals.

Good luck with your research! And if we can be of any help, please do
not hesitate to contact us.

Marja van Aken
Scientific Software Developer
SVI, makers of the Huygens software
[hidden email]



On 07/23/2012 09:15 PM, Krzysztof Berniak wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> *****
>
> Dear List Members,
>
> We are looking for an image analysis/cytometry tool that can perform
> the following job:
>
> We have a 3D image stack that consists of several hundred green and
> red foci (they represent two nuclear processes that occur in many
> little spots). We need to find the center of mass of each red 3D
> focus, and measure the distance (in 3D) to the nearest green spot, and
> repeat the process for all red spots in one 3D stack. Next, we need to
> repeat this job in a large number of nuclei. Eventually, we need to
> obtain a histogram of distances between the red spots and their the
> nearest green spot neighbor in a large population of 3D objects.
>
> Does anyone know a ready-to-use image analysis tool that can do this
> job for us? We have started to work on our own algorithm, but maybe
> there are solutions to this problem available already. Any advice
> about a software package that could this work would be much
> appreciated.
>
> Thank you very much for your help,
>
> Krzysztof Berniak, Jurek Dobrucki
> Jagiellonian University
>