Cells Tracking (and intensity plot)

[adrienlibre]

Hello,
I am new to ICY and I find it is a very nice program with many potential and which offers many possibilities. So thank you for that :)
However, I still didn't manage to understand how to properly track cells, and I would be glad if you could help me.

In particular, I would like to achieve the following steps:
1) To manage to track properly a single cell, following its contour (or a mask based on threshold) over time.
2) To be able to track all cells at the same time.
3) To be able to plot, and to export, the average intensity of each cell (internal to each contour or internal to each mask) over time.

I don't mind if cells segmentation is not totally automatically.
In fact, if I could be able to use the first frame to manually segment cells, and then obtain a tracking of each cell based on my first help, with the possibility of manually correcting errors and to add tracks after divisions, it would be great.
I tried to use the active contours tool and from what I understand it allows this option, but I didn't manage to find good results nor I found a tutorial to understand the plugin properly.

I think that a forum answer (and, some day, a tutorial) with a resolution step by step of a task like this one, would be greatly useful for many users.
I attached a tif file with a sequence of 10 frames, which corresponds to the problem I have to solve.

Thank you very much, Adriano

[Arthur Yakimovich]

Hey Guys,


Same here. I just installed Icy and find it very nice, but I failed to find documentation on how to do segmentation and tracking of cell nuclei in an automatic way.
Would appreciate any help. thanks a lot in advance.

[Fab]

Hi !

I just created a video tutorial for you:

http://www.youtube.com/watch?v=APewAXxMkJE&feature=youtu.be

I hope this will help you

Best,

Fabrice

[adrienlibre]

Dear Fabrice, 

thank you, this is a first help. But actually, in general I would like to be able to do something more complex:

1) Detected spots are sometimes jumping up from nothing, or merging, or following the wrong path, isn't it possible delete some of them, and correct connections of spots between frames?

2) In this way you are not really segmenting cells, but detecting spots. And this implies various problems. One of those is, I am detecting intensity of the spot, while I want average intensity of the cell.

In general, it would be great for me, if you could explain me how to obtain a semi-automatic segmentation, (that means for me, I can correct it when it is wrong), maybe through active contour, and then plot the average intensity of each one of the segmented cells (of course, always performing tracking) .

I am sure this is possible, it would be of great use for many of us a tutorial or some verbal explanation also on that issue.

Thank you again, merci!

Adriano

[Fab]

Hi !

1 / It's possible to correct the generated tracks via the trackmanager. You can have a look at the end of the documentation here: http://icy.bioimageanalysis.org/plugin/Track_Manager
At the moment you can correct within the trackmanager but not directly over the canvas, it's something I should do to smooth its use.

2/ If you enable the binary output in the spotdector > detector, you will have a look to what is actually segmented. I guess it matches with what you wish. Still for exact contours you can use either k-means or active contours. But I should double check if k-means exports detections that the tracker can manage. Maybe Alex can tell us.

about your last question: the program provides min/max/average and std deviation around the barycenter considering a diameter. In the video, you can see I set this parameter just before export to xls.

Hope this helps,

Fabrice

[adrienlibre]

Dear Fabrice,
as a first thing, thank you for your kindness.

Regarding option 1/, since many times spots are generated or merged or detected in a wrong way, tracks come out very messy. I tried correcting tracks though trackmanager, but this is quite difficult. A great, great improvement, would be to integrate on your plugin the same track creation mechanism of the wonderful plugin Speckle TrackerJ plugin for ImageJ, in which one can correct, or even create the track, simply clicking on subsequent frames.

Here are Speckle TrackerJ site and reference:
http://athena.physics.lehigh.edu/speckletrackerj/basic.chy
Reference: M.B. Smith¹, E. Karatekin^2,3, A. Gohlke³, H. Mizuno⁴, N. Watanabe³, and D. Vavylonis¹, Biophys J, 101:1794-1804 (2011)

The only thing to add to this plugin in order to have a state-of-the art cell tracker would be then the automatic detection of cell contours once cells are (manually or semiautomatically) "spotted".


Regarding option 2/, I tried to obtain cells tracking starting from "active contours", but because of the lack of a tutorial I didn't manage to obtain a proper multicell tracking. I noticed some nice features, as the possibility to change active contour shape once created, but in the end I never manage to grasp the exact way on how to do it.
If possible, a tutorial on how to achieve cell tracking through active contours would also be great.

I think that if it will be possible to achieve a good and fast semiautomatic cell tracking plugin (or a very robust and good tutorial on how to achieve this purpose with the already existing Icy Plugins) this will be of great use for many people.
If you want, we can also have a more extensive chat via telephone or skype.
I could then give back to the community by writing a clear and comprehensive tutorial once we'll figure out how to solve the issue.

Thank you,
Adriano

[Alexandre Dufour]

Hi Adriano,

Sorry I didn't find the time to step in the discussion earlier.

Regarding option 2/, I tried to obtain cells tracking starting from "active contours", but because of the lack of a tutorial I didn't manage to obtain a proper multicell tracking. I noticed some nice features, as the possibility to change active contour shape once created, but in the end I never manage to grasp the exact way on how to do it.
If possible, a tutorial on how to achieve cell tracking through active contours would also be great.

I did try the active contours on your test sequence, however since the daughter cell remain very close to the mother cell (I assume this is a division we are observing ?), the contour never manages to properly split and therefore detect two distinct cells.

Regarding the tutorial, as discussed in other posts I am working on a new version of the tool that will be simpler to use (less parameters, lighter interface). It will also be more tailored toward the use of Protocols, in which I expect to allow more flexibility onto how the cells are detected and tracked.

I think that if it will be possible to achieve a good and fast semiautomatic cell tracking plugin (or a very robust and good tutorial on how to achieve this purpose with the already existing Icy Plugins) this will be of great use for many people.

If you know the literature well, you will acknowledge that cell tracking is not a simple problem with a potentially unique answer. Indeed the rationale may differ from a dataset to the next (for instance, I already use the active contours plugin to perform automated cell tracking on many datasets, I just never had the chance to work on images such as yours).

Together with your tutorial idea, we hope to solve such problems via protocols, i.e., by publishing protocols tailored to various cases, hoping that the user will select the one that suits most his/her case and tweak it from there.

Any further input is most welcome, thanks for joining in !

Alexandre

[Fab]

Dear Adriano,

Regarding option 1/, since many times spots are generated or merged or detected in a wrong way, tracks come out very messy. I tried correcting tracks though trackmanager, but this is quite difficult. A great, great improvement, would be to integrate on your plugin the same track creation mechanism of the wonderful plugin Speckle TrackerJ plugin for ImageJ, in which one can correct, or even create the track, simply clicking on subsequent frames.

Here are Speckle TrackerJ site and reference:
http://athena.physics.lehigh.edu/speckletrackerj/basic.chy
Reference: M.B. Smith¹, E. Karatekin^2,3, A. Gohlke³, H. Mizuno⁴, N. Watanabe³, and D. Vavylonis¹, Biophys J, 101:1794-1804 (2011)

The only thing to add to this plugin in order to have a state-of-the art cell tracker would be then the automatic detection of cell contours once cells are (manually or semiautomatically) "spotted".

Thank you for your feedback. I will have a closer look at the plugin you mention, and try to implement this in a future version of the trackmanager.

I also like you idea of tutorial feedback ! I should arrange some blog or tutorial area over the Icy website to welcome this kind of initiative !

Best,

Fabrice