Fwd: [fiji-devel] Re: tracking algorithms?

[lechristophe]

---------- Forwarded message ----------
From: Christophe Leterrier <[hidden email]>
Date: 2010/8/9
Subject: Re: [fiji-devel] Re: tracking algorithms?
To: Nick Perry <[hidden email]>


I'm not an image processing specialist, but I'm following the
applications of particle tracking in biology.

I used the program from A. Serge and N. Berteaux to track quantum
dots-labeled surface particles, and it worked quite well:

Sergé A, Bertaux N, Rigneault H, Marguet D.
Dynamic multiple-target tracing to probe spatiotemporal cartography of
cell membranes.
Nat Methods. 2008 Aug;5(8):687-94
Single-particle tracking (SPT) of quantum-dots has been a proeminent
application of tracking algorithms for several years now, with the
added complexity that these molecules are blinking, so an individual
QD can disappear from several frames before reappearing in the
subsequent ones.

In the same issue of Nature Methods, I remember reading another paper
from G. Danuser's group that was rather complicated for my
mathematically-challenged skills, but claiming to solve the 2D+t
trajectories assignment problem as a whole:

Jaqaman K, Loerke D, Mettlen M, Kuwata H, Grinstein S, Schmid SL, Danuser G.
Robust single-particle tracking in live-cell time-lapse sequences.
Nat Methods. 2008 Aug;5(8):695-702.

There is also a nice review from E. Meijering that you can download here:
http://www.imagescience.org/meijering/publications/?p=1001

Finally, several recent short communications in Nature Methods have
adressed the problem of the most efficient and fast ways of detecting
and localizing spots on images (in order to speed up superresolutive
imaging modalities based on localization of single fluorescent
emitters). This could be useful for the first steps of tracking.


Christophe Leterrier


2010/8/9 Nick Perry <[hidden email]>

>
> Hi Wilhelm,
> I hadn't looked at that. Thanks for the tip!
> Nick
>
> On Mon, Aug 9, 2010 at 8:28 AM, Wilhelm Burger <[hidden email]> wrote:
>>
>> Hi Nicolas,
>>
>> did you look at the Image_Stabilizer plugin listed on
>> http://rsbweb.nih.gov/ij/plugins/index.html (http://www.kangli.org/
>> code/Image_Stabilizer.html)? It is apparently based on the Lucas-
>> Kanade tracker, which is a commonly used standard algorithm (at least
>> for 2D tracking).
>>
>> --Wilhelm
>>
>>
>> On 8 Aug., 21:38, Nick Perry <[hidden email]> wrote:
>> > Hi all,
>> >
>> > *My goal:*
>> > I'm about to write a tracking class, which tracks points over multiple time
>> > frames in 3D (or 2D). I looked on the Fiji website and it looks like there
>> > really only exists a 2D tracker, a manual tracker, and ToAST (which also
>> > seems to be 2D over time, if the plugin hasn't been significantly modified
>> > from the paper describing it).
>> >
>> > Does anyone have any suggestions for potential, robust-as-possible
>> > algorithms? Ideally it should be able to handle a few particles, to many
>> > hundreds, and be generally applicable in terms of application areas. I am
>> > mostly interested in either: (a) something that has already been written by
>> > Fiji developers (like the extrema finding class I hadn't known about); or
>> > (b) if there is some prevailing algorithm out there that I haven't read
>> > about that would make this more useful to others. Ideally, I'd like to make
>> > an algorithm that other people *want* use, otherwise I find it to be a waste
>> > of time. So it should be able to do the things people here generally would
>> > want out of a tracker (as much as possible).
>> >
>> > *My own thoughts, approaches:*
>> > I've done some research of my own, and specifically focused on the following
>> > algorithms:
>> >
>> > - "StarryNite," Bao et al. 2006 "Automated cell lineage tracing in
>> > Caenorhabditis elegans" (
>> > - Sbalzarini and Koumoutsakos, 2005 "Feature point tracking and trajectory
>> > analysis for video imaging in cell biology"
>> > - A review by Y. Kalaidzidis, 2007 "Intracellular objects tracking"
>> > - IPAN, Verestoy and Chetverikov, 1998 "Tracking feature points: A new
>> > algorithm"
>> >
>> > I think if I make an algorithm, it would be more of a combination of
>> > StarryNite, and the algorithm described by Sbalzarini and Koumousakos.
>> >
>> > Specifically, I would use a greedy approach to try to pick the best track
>> > "connections," and my greedy criterion would actually be a combination of
>> > many aspects of the objects being tracked, including:
>> >
>> > - the position of the object in frame t+1 is "similar enough" to frame t.
>> > - how similar the potential object in frame t+1 looks compared to the object
>> > in frame t (intensity, size, shape)
>> >
>> > The plugin should also allow for entry/exit of objects, as well as allow
>> > developers to create additional classes that incorporate application
>> > specific information into the track assignments. For example, in the case of
>> > C. elegans development, I would try to create a system which allows for a
>> > user to incorporate information on the shape of the objects (round,
>> > condense, or elongated, which should match certain timing criterion in order
>> > to be valid) to more intelligently assign tracks.
>> >
>> > Again, these are some of my ideas, and if any of you have any input on other
>> > algorithms I should take a look at, or other aspects of tracking you think I
>> > should include in my algorithm, please let me know! It would be much
>> > appreciated!
>> >
>> > Cheers,
>> > Nick
>> >
>> > --
>> > [hidden email]
>> > 425.308.2550
>> > @nick_perry
>>
>> --
>> You received this message because you are subscribed to the Google Groups "Fiji-devel" group.
>> To post to this group, send email to [hidden email].
>> To unsubscribe from this group, send email to [hidden email].
>> For more options, visit this group at http://groups.google.com/group/fiji-devel?hl=en.
>>
>
>
>
> --
> [hidden email]
> 425.308.2550
> @nick_perry
>
> --
> You received this message because you are subscribed to the Google Groups "Fiji-devel" group.
> To post to this group, send email to [hidden email].
> To unsubscribe from this group, send email to [hidden email].
> For more options, visit this group at http://groups.google.com/group/fiji-devel?hl=en.

[lechristophe]

Sorry for this, I mistakingly forwarded a fiji-devel post to the
ImageJ mailing list.

Christophe

2010/8/9 Christophe Leterrier <[hidden email]>:

> ---------- Forwarded message ----------
> From: Christophe Leterrier <[hidden email]>
> Date: 2010/8/9
> Subject: Re: [fiji-devel] Re: tracking algorithms?
> To: Nick Perry <[hidden email]>
>
>
> I'm not an image processing specialist, but I'm following the
> applications of particle tracking in biology.
>
> I used the program from A. Serge and N. Berteaux to track quantum
> dots-labeled surface particles, and it worked quite well:
>
> Sergé A, Bertaux N, Rigneault H, Marguet D.
> Dynamic multiple-target tracing to probe spatiotemporal cartography of
> cell membranes.
> Nat Methods. 2008 Aug;5(8):687-94
> Single-particle tracking (SPT) of quantum-dots has been a proeminent
> application of tracking algorithms for several years now, with the
> added complexity that these molecules are blinking, so an individual
> QD can disappear from several frames before reappearing in the
> subsequent ones.
>
> In the same issue of Nature Methods, I remember reading another paper
> from G. Danuser's group that was rather complicated for my
> mathematically-challenged skills, but claiming to solve the 2D+t
> trajectories assignment problem as a whole:
>
> Jaqaman K, Loerke D, Mettlen M, Kuwata H, Grinstein S, Schmid SL, Danuser G.
> Robust single-particle tracking in live-cell time-lapse sequences.
> Nat Methods. 2008 Aug;5(8):695-702.
>
> There is also a nice review from E. Meijering that you can download here:
> http://www.imagescience.org/meijering/publications/?p=1001
>
> Finally, several recent short communications in Nature Methods have
> adressed the problem of the most efficient and fast ways of detecting
> and localizing spots on images (in order to speed up superresolutive
> imaging modalities based on localization of single fluorescent
> emitters). This could be useful for the first steps of tracking.
>
>
> Christophe Leterrier
>
>
> 2010/8/9 Nick Perry <[hidden email]>
>>
>> Hi Wilhelm,
>> I hadn't looked at that. Thanks for the tip!
>> Nick
>>
>> On Mon, Aug 9, 2010 at 8:28 AM, Wilhelm Burger <[hidden email]> wrote:
>>>
>>> Hi Nicolas,
>>>
>>> did you look at the Image_Stabilizer plugin listed on
>>> http://rsbweb.nih.gov/ij/plugins/index.html (http://www.kangli.org/
>>> code/Image_Stabilizer.html)? It is apparently based on the Lucas-
>>> Kanade tracker, which is a commonly used standard algorithm (at least
>>> for 2D tracking).
>>>
>>> --Wilhelm
>>>
>>>
>>> On 8 Aug., 21:38, Nick Perry <[hidden email]> wrote:
>>> > Hi all,
>>> >
>>> > *My goal:*
>>> > I'm about to write a tracking class, which tracks points over multiple time
>>> > frames in 3D (or 2D). I looked on the Fiji website and it looks like there
>>> > really only exists a 2D tracker, a manual tracker, and ToAST (which also
>>> > seems to be 2D over time, if the plugin hasn't been significantly modified
>>> > from the paper describing it).
>>> >
>>> > Does anyone have any suggestions for potential, robust-as-possible
>>> > algorithms? Ideally it should be able to handle a few particles, to many
>>> > hundreds, and be generally applicable in terms of application areas. I am
>>> > mostly interested in either: (a) something that has already been written by
>>> > Fiji developers (like the extrema finding class I hadn't known about); or
>>> > (b) if there is some prevailing algorithm out there that I haven't read
>>> > about that would make this more useful to others. Ideally, I'd like to make
>>> > an algorithm that other people *want* use, otherwise I find it to be a waste
>>> > of time. So it should be able to do the things people here generally would
>>> > want out of a tracker (as much as possible).
>>> >
>>> > *My own thoughts, approaches:*
>>> > I've done some research of my own, and specifically focused on the following
>>> > algorithms:
>>> >
>>> > - "StarryNite," Bao et al. 2006 "Automated cell lineage tracing in
>>> > Caenorhabditis elegans" (
>>> > - Sbalzarini and Koumoutsakos, 2005 "Feature point tracking and trajectory
>>> > analysis for video imaging in cell biology"
>>> > - A review by Y. Kalaidzidis, 2007 "Intracellular objects tracking"
>>> > - IPAN, Verestoy and Chetverikov, 1998 "Tracking feature points: A new
>>> > algorithm"
>>> >
>>> > I think if I make an algorithm, it would be more of a combination of
>>> > StarryNite, and the algorithm described by Sbalzarini and Koumousakos.
>>> >
>>> > Specifically, I would use a greedy approach to try to pick the best track
>>> > "connections," and my greedy criterion would actually be a combination of
>>> > many aspects of the objects being tracked, including:
>>> >
>>> > - the position of the object in frame t+1 is "similar enough" to frame t.
>>> > - how similar the potential object in frame t+1 looks compared to the object
>>> > in frame t (intensity, size, shape)
>>> >
>>> > The plugin should also allow for entry/exit of objects, as well as allow
>>> > developers to create additional classes that incorporate application
>>> > specific information into the track assignments. For example, in the case of
>>> > C. elegans development, I would try to create a system which allows for a
>>> > user to incorporate information on the shape of the objects (round,
>>> > condense, or elongated, which should match certain timing criterion in order
>>> > to be valid) to more intelligently assign tracks.
>>> >
>>> > Again, these are some of my ideas, and if any of you have any input on other
>>> > algorithms I should take a look at, or other aspects of tracking you think I
>>> > should include in my algorithm, please let me know! It would be much
>>> > appreciated!
>>> >
>>> > Cheers,
>>> > Nick
>>> >
>>> > --
>>> > [hidden email]
>>> > 425.308.2550
>>> > @nick_perry
>>>
>>> --
>>> You received this message because you are subscribed to the Google Groups "Fiji-devel" group.
>>> To post to this group, send email to [hidden email].
>>> To unsubscribe from this group, send email to [hidden email].
>>> For more options, visit this group at http://groups.google.com/group/fiji-devel?hl=en.
>>>
>>
>>
>>
>> --
>> [hidden email]
>> 425.308.2550
>> @nick_perry
>>
>> --
>> You received this message because you are subscribed to the Google Groups "Fiji-devel" group.
>> To post to this group, send email to [hidden email].
>> To unsubscribe from this group, send email to [hidden email].
>> For more options, visit this group at http://groups.google.com/group/fiji-devel?hl=en.
>

[chris elliott]

Thanks, I had not seen it, but like the mtrack2 plugin for following drosophila larvae

http://biolpc22.york.ac.uk/drosophila/larvae/


chris

On 9 Aug 2010, at 14:50, Christophe Leterrier wrote:

> Sorry for this, I mistakingly forwarded a fiji-devel post to the
> ImageJ mailing list.
>
> Christophe
>
> 2010/8/9 Christophe Leterrier <[hidden email]>:
>> ---------- Forwarded message ----------
>> From: Christophe Leterrier <[hidden email]>
>> Date: 2010/8/9
>> Subject: Re: [fiji-devel] Re: tracking algorithms?
>> To: Nick Perry <[hidden email]>
>>
>>
>> I'm not an image processing specialist, but I'm following the
>> applications of particle tracking in biology.
>>
>> I used the program from A. Serge and N. Berteaux to track quantum
>> dots-labeled surface particles, and it worked quite well:
>>
>> Sergé A, Bertaux N, Rigneault H, Marguet D.
>> Dynamic multiple-target tracing to probe spatiotemporal cartography of
>> cell membranes.
>> Nat Methods. 2008 Aug;5(8):687-94
>> Single-particle tracking (SPT) of quantum-dots has been a proeminent
>> application of tracking algorithms for several years now, with the
>> added complexity that these molecules are blinking, so an individual
>> QD can disappear from several frames before reappearing in the
>> subsequent ones.
>>
>> In the same issue of Nature Methods, I remember reading another paper
>> from G. Danuser's group that was rather complicated for my
>> mathematically-challenged skills, but claiming to solve the 2D+t
>> trajectories assignment problem as a whole:
>>
>> Jaqaman K, Loerke D, Mettlen M, Kuwata H, Grinstein S, Schmid SL, Danuser G.
>> Robust single-particle tracking in live-cell time-lapse sequences.
>> Nat Methods. 2008 Aug;5(8):695-702.
>>
>> There is also a nice review from E. Meijering that you can download here:
>> http://www.imagescience.org/meijering/publications/?p=1001
>>
>> Finally, several recent short communications in Nature Methods have
>> adressed the problem of the most efficient and fast ways of detecting
>> and localizing spots on images (in order to speed up superresolutive
>> imaging modalities based on localization of single fluorescent
>> emitters). This could be useful for the first steps of tracking.
>>
>>
>> Christophe Leterrier
>>
>>
>> 2010/8/9 Nick Perry <[hidden email]>
>>>
>>> Hi Wilhelm,
>>> I hadn't looked at that. Thanks for the tip!
>>> Nick
>>>
>>> On Mon, Aug 9, 2010 at 8:28 AM, Wilhelm Burger <[hidden email]> wrote:
>>>>
>>>> Hi Nicolas,
>>>>
>>>> did you look at the Image_Stabilizer plugin listed on
>>>> http://rsbweb.nih.gov/ij/plugins/index.html (http://www.kangli.org/
>>>> code/Image_Stabilizer.html)? It is apparently based on the Lucas-
>>>> Kanade tracker, which is a commonly used standard algorithm (at least
>>>> for 2D tracking).
>>>>
>>>> --Wilhelm
>>>>
>>>>
>>>> On 8 Aug., 21:38, Nick Perry <[hidden email]> wrote:
>>>>> Hi all,
>>>>>
>>>>> *My goal:*
>>>>> I'm about to write a tracking class, which tracks points over multiple time
>>>>> frames in 3D (or 2D). I looked on the Fiji website and it looks like there
>>>>> really only exists a 2D tracker, a manual tracker, and ToAST (which also
>>>>> seems to be 2D over time, if the plugin hasn't been significantly modified
>>>>> from the paper describing it).
>>>>>
>>>>> Does anyone have any suggestions for potential, robust-as-possible
>>>>> algorithms? Ideally it should be able to handle a few particles, to many
>>>>> hundreds, and be generally applicable in terms of application areas. I am
>>>>> mostly interested in either: (a) something that has already been written by
>>>>> Fiji developers (like the extrema finding class I hadn't known about); or
>>>>> (b) if there is some prevailing algorithm out there that I haven't read
>>>>> about that would make this more useful to others. Ideally, I'd like to make
>>>>> an algorithm that other people *want* use, otherwise I find it to be a waste
>>>>> of time. So it should be able to do the things people here generally would
>>>>> want out of a tracker (as much as possible).
>>>>>
>>>>> *My own thoughts, approaches:*
>>>>> I've done some research of my own, and specifically focused on the following
>>>>> algorithms:
>>>>>
>>>>> - "StarryNite," Bao et al. 2006 "Automated cell lineage tracing in
>>>>> Caenorhabditis elegans" (
>>>>> - Sbalzarini and Koumoutsakos, 2005 "Feature point tracking and trajectory
>>>>> analysis for video imaging in cell biology"
>>>>> - A review by Y. Kalaidzidis, 2007 "Intracellular objects tracking"
>>>>> - IPAN, Verestoy and Chetverikov, 1998 "Tracking feature points: A new
>>>>> algorithm"
>>>>>
>>>>> I think if I make an algorithm, it would be more of a combination of
>>>>> StarryNite, and the algorithm described by Sbalzarini and Koumousakos.
>>>>>
>>>>> Specifically, I would use a greedy approach to try to pick the best track
>>>>> "connections," and my greedy criterion would actually be a combination of
>>>>> many aspects of the objects being tracked, including:
>>>>>
>>>>> - the position of the object in frame t+1 is "similar enough" to frame t.
>>>>> - how similar the potential object in frame t+1 looks compared to the object
>>>>> in frame t (intensity, size, shape)
>>>>>
>>>>> The plugin should also allow for entry/exit of objects, as well as allow
>>>>> developers to create additional classes that incorporate application
>>>>> specific information into the track assignments. For example, in the case of
>>>>> C. elegans development, I would try to create a system which allows for a
>>>>> user to incorporate information on the shape of the objects (round,
>>>>> condense, or elongated, which should match certain timing criterion in order
>>>>> to be valid) to more intelligently assign tracks.
>>>>>
>>>>> Again, these are some of my ideas, and if any of you have any input on other
>>>>> algorithms I should take a look at, or other aspects of tracking you think I
>>>>> should include in my algorithm, please let me know! It would be much
>>>>> appreciated!
>>>>>
>>>>> Cheers,
>>>>> Nick
>>>>>
>>>>> --
>>>>> [hidden email]
>>>>> 425.308.2550
>>>>> @nick_perry
>>>>
>>>> --
>>>> You received this message because you are subscribed to the Google Groups "Fiji-devel" group.
>>>> To post to this group, send email to [hidden email].
>>>> To unsubscribe from this group, send email to [hidden email].
>>>> For more options, visit this group at http://groups.google.com/group/fiji-devel?hl=en.
>>>>
>>>
>>>
>>>
>>> --
>>> [hidden email]
>>> 425.308.2550
>>> @nick_perry
>>>
>>> --
>>> You received this message because you are subscribed to the Google Groups "Fiji-devel" group.
>>> To post to this group, send email to [hidden email].
>>> To unsubscribe from this group, send email to [hidden email].
>>> For more options, visit this group at http://groups.google.com/group/fiji-devel?hl=en.
>>