Re: track multiple ROIs movements

[dpoburko]

Hi All,

  As usual, I apologize is I missed some previous threads on this topic.

  I am trying to track multiple (50 - ~500) ROIs over time. Basically,
every few minutes, I grab a reference image that will be segmented to
define ROIs for extracting dynamic responses from a different channel
that is acquired with higher temporal resolution. So essentially, I need
a method to compare and register ROIs in reference images (in this case
a synaptic bouton) over time. At the moment, I am simply calculating a
minimum distance between ROI centroids in Excel, and excluding those
that are above a cut-off distance. I am wondering if anyone has
experience with or can suggest a more sophisticated method. For example,
something that takes into account the relative position of one ROI (or
centroid) in a constellation of numerous other centroids. Or perhaps
analyze the distance and direction to several nearest neighbors, and use
that a criteria to compare with the next image. My plan is eventually to
run this in a form of batch mode, so it doesn't have to be terribly fast.  

Many thanks,
Damon
--

Damon Poburko, PhD
Postdoctoral Research Fellow
Stanford University School of Medicine
Dept. of Molecular & Cellular Physiology
279 Campus Dr., Beckman B103, Stanford, CA 94305
Ph: 650 725 7564, fax: [phone]

[David Webster]

Damon,

I have done this type of point pattern matching (or alignment) using a
Generalized Hough Transform. For what it's worth, this is nothng like the
regular Hough Transform used to look for lines or circles.

David

On Mon, Apr 20, 2009 at 10:20 AM, Damon Poburko <[hidden email]>wrote:

> Hi All,
>
>  As usual, I apologize is I missed some previous threads on this topic.
>
>  I am trying to track multiple (50 - ~500) ROIs over time. Basically, every
> few minutes, I grab a reference image that will be segmented to define ROIs
> for extracting dynamic responses from a different channel that is acquired
> with higher temporal resolution. So essentially, I need a method to compare
> and register ROIs in reference images (in this case a synaptic bouton) over
> time. At the moment, I am simply calculating a minimum distance between ROI
> centroids in Excel, and excluding those that are above a cut-off distance. I
> am wondering if anyone has experience with or can suggest a more
> sophisticated method. For example, something that takes into account the
> relative position of one ROI (or centroid) in a constellation of numerous
> other centroids. Or perhaps analyze the distance and direction to several
> nearest neighbors, and use that a criteria to compare with the next image.
> My plan is eventually to run this in a form of batch mode, so it doesn't
> have to be terribly fast.
> Many thanks,
> Damon
> --
>
> Damon Poburko, PhD
> Postdoctoral Research Fellow
> Stanford University School of Medicine
> Dept. of Molecular & Cellular Physiology
> 279 Campus Dr., Beckman B103, Stanford, CA 94305
> Ph: 650 725 7564, fax: [phone]
>

[Michael Schmid]

Hi Damon,

if you convert the ROI centers to images with a single dot there, you  
could try the Particle Tracker
   http://weeman.inf.ethz.ch/particletracker/

The easiest way to do the conversion to single dots:
     edit>selection>create mask,
   thereafter
     if you have nice convex ROIs: process>binary>ultimate points &  
apply threshold
     otherwise process>binary>find maxima with single-point output
Of course, you can also draw dots in a new image.

Michael
________________________________________________________________

On 20 Apr 2009, at 19:20, Damon Poburko wrote:

> Hi All,
>
>  As usual, I apologize is I missed some previous threads on this  
> topic.
>
>  I am trying to track multiple (50 - ~500) ROIs over time.  
> Basically, every few minutes, I grab a reference image that will be  
> segmented to define ROIs for extracting dynamic responses from a  
> different channel that is acquired with higher temporal resolution.  
> So essentially, I need a method to compare and register ROIs in  
> reference images (in this case a synaptic bouton) over time. At the  
> moment, I am simply calculating a minimum distance between ROI  
> centroids in Excel, and excluding those that are above a cut-off  
> distance. I am wondering if anyone has experience with or can  
> suggest a more sophisticated method. For example, something that  
> takes into account the relative position of one ROI (or centroid)  
> in a constellation of numerous other centroids. Or perhaps analyze  
> the distance and direction to several nearest neighbors, and use  
> that a criteria to compare with the next image. My plan is  
> eventually to run this in a form of batch mode, so it doesn't have  
> to be terribly fast.
> Many thanks,
> Damon
> --
>
> Damon Poburko, PhD
> Postdoctoral Research Fellow
> Stanford University School of Medicine
> Dept. of Molecular & Cellular Physiology
> 279 Campus Dr., Beckman B103, Stanford, CA 94305
> Ph: 650 725 7564, fax: [phone]