Some sampling and detection questions on Active Contour, Wavelet Spot Detector and ROI Inclusion analysis.

[Yasmine Khalil]

Hi everybody,

I have a lot of quite technical questions of sampling, detection and segmentation. Could someone enlighten me?

1- When I run an Active Contour to segment cells, how are the center of mass and the radius of the cell calculated? Are they computed by the Active Contour bloc or are they calculated after that by ROI Inclusion analysis?
And how is the center of mass calculated? Is it simply a weighted (or non-wieghted) barycenter of the cell mask?

2- I have the same question for the spots's position. I detect spots with Wavelet Spot Detector. How is the center of mass of a spot estimated? And what is the potential error of estimation? I ask the question because it seemed to be the barycenter (weighted or not?) of the "occupied" pixels...

3- When I have spots inside my cell, ROI Inclusion analysis gives me Oriented distance to edge in "pixels" and in "%" . So if I zoom, I have locally something like in the image I attached here.  How is the distance calculated? And when it's normalized with the radius, how is this radius obtained?

4- And finally, when Icy passes from HK-Means bloc to Active contour bloc. Which intensity class is used as initial condition for the segmentation algorithme? Is it the brightest class or the darker one? (In other words, does the segmentation begin outside or inside the cells?) And is it possible to decide about it?


It would be great if you could enlighten me because I need to know the error that could be generated by my protocol.

Thanks a lot.

Yasmine.

[Yasmine Khalil]

Hmmm...nothing?

[Stephane]

Hi Yasmine,

I'm willing to reply your questions but definitely i'm not the best placed for that (i think Alexandre is as he wrote most of these plugins). At least i can reply about the mass center : I think they all use the barycenter of ROI mask to compute it (as does Icy in the ROI panel).

Best,

- Stephane

[Alexandre Dufour]

Some (brief answers) below. I however would suggest that you split your questions on the forum in the future, for the benefit of other users (they may not use your entire workflow, but just a piece of it, so splitting makes finding answer easier).

On 10 May 2017, at 14:48, Yasmine Khalil <khalil.y...@gmail.com> wrote:

1- When I run an Active Contour to segment cells, how are the center of mass and the radius of the cell calculated? Are they computed by the Active Contour bloc or are they calculated after that by ROI Inclusion analysis?
And how is the center of mass calculated? Is it simply a weighted (or non-wieghted) barycenter of the cell mask?

This is calculated within ROI inclusion, and indeed is the barycentre of the ROI mask (therefore a decimal value for each coordinate).

2- I have the same question for the spots's position. I detect spots with Wavelet Spot Detector. How is the center of mass of a spot estimated? And what is the potential error of estimation? I ask the question because it seemed to be the barycenter (weighted or not?) of the "occupied" pixels…

Same method indeed

3- When I have spots inside my cell, ROI Inclusion analysis gives me Oriented distance to edge in "pixels" and in "%" . So if I zoom, I have locally something like in the image I attached here.  How is the distance calculated? And when it's normalized with the radius, how is this radius obtained?

All distances (including radius) are measured from the barycentre (or the cell or spot, see above) to the furthest pixel (technically, its top left corner, if you want this level of detail). However a “better” answer to this question is rather a warning: at this level of detail (i.e. under a pixel), small variations in distance/percentage should not be relied on, as they are clearly negligible compared to various experimental artefacts affecting image resolution (camera sampling, microscope PSF, chromatic aberrations, just to name a few). 

4- And finally, when Icy passes from HK-Means bloc to Active contour bloc. Which intensity class is used as initial condition for the segmentation algorithme? Is it the brightest class or the darker one? (In other words, does the segmentation begin outside or inside the cells?) And is it possible to decide about it?

Technically speaking, the “darkest” one, although the definition of “darkest” depends on each object. The HK-Means browses bottom-up (i.e. from darkest to brightest) and extracts all objects that match the size range in the current class (and stops there for these objects). So each object may have been initialised thanks to a different class (that’s actually the whole point of the algorithm).

Alexandre

[Yasmine Khalil]

Thank you Alexandre, you cleared it up a lot! :D

Next time I will seperate my questions, it's understood.

Thanks a lot.