Active contour and 3D active mesh plugins

[louise-laure]

Hello Alexandre and Icy users,

I'm trying to use Active contour and 3D active mesh plugins to analyse images of cells loaded with various Fluo dyes in 3D over time.

For 1 dye, I have managed with Alexandre's protocol (cf post: "How to measure volume and suface of a cell filmed in a .tiff 4-d stack)
to set the parameters of 3D active mesh plugin to find my cells and follow them over time.
We're now missing an Intensity option in the TrackProcessor Mesh Analysis plugin but this will come out soon I hope :-)
Btwy, when I run it, appart from finding my cells and tracks, it also gives me another image .bin with is just a blank image of the same numbers of z and t frames... has it happened to anyone using 3D active mesh ?

For dye n°2, (Ca imaging)
I have now another issue,
which is due to the fact that some of my cells only seem to appear at some precise time point when they fire an Action potential.
Even if I tried to set the previous protocol block "extract time" to the time position where I can see them, it seems to have a hard time finding them.
The problem may also be related to the fact that these 3D images are of only z=3 or 4, hence the Gaussian blur can't be done properly to have a good detection by HK means because the z blur is limited to 1.3 max...

So I've tried to use active contour with a 2D version of my images but I don't seem to work it well because it gives me a new image with no detections...
I've tried loading the file of the 2D version but seems too big.
So I sent it directly to your e-mail Alexandre.

Thanks by advance for the help
LL

[Alexandre Dufour]

Hi LL,

On 5 nov. 2013, at 15:57, louise-laure wrote:

> Hello Alexandre and Icy users,
>
> I'm trying to use Active contour and 3D active mesh plugins to analyse images of cells loaded with various Fluo dyes in 3D over time.
>
> For 1 dye, I have managed with Alexandre's protocol (cf post: "How to measure volume and suface of a cell filmed in a .tiff 4-d stack)
> to set the parameters of 3D active mesh plugin to find my cells and follow them over time.
> We're now missing an Intensity option in the TrackProcessor Mesh Analysis plugin but this will come out soon I hope :-)

Yes, TODO list is locked and loaded :)

> Btwy, when I run it, appart from finding my cells and tracks, it also gives me another image .bin with is just a blank image of the same numbers of z and t frames... has it happened to anyone using 3D active mesh ?

The image is probably not blank, it should be a labeled sequence (of same size as the original) in which each object is filled with its id (i.e. a number from 1 to "n", where "n" is probably small compared to the max possible value of about 2 billion). So it's just a contrast issue: if you scroll the red contrast bar to the left, you'll see the labels magically appear :)

> For dye n°2, (Ca imaging)
> I have now another issue,
> which is due to the fact that some of my cells only seem to appear at some precise time point when they fire an Action potential.
> Even if I tried to set the previous protocol block "extract time" to the time position where I can see them, it seems to have a hard time finding them.
> The problem may also be related to the fact that these 3D images are of only z=3 or 4, hence the Gaussian blur can't be done properly to have a good detection by HK means because the z blur is limited to 1.3 max...

Indeed the very few z sections you have are most probably the reason. In fact, your objects might not even look "3D" although they are... there's not much out of that one...

For your own data, I'll reply off-list, as the data itself seems corrupted.

Alexandre

[louise-laure]

Hi Alexandre and everyone,
thanks,

indeed the histogram is weird,
but I am wondering how it could happen,
since I acquired with exactly the same way using the same hardware (same 2-photon microscope, Hybrid detectors and same sofware Leica),
and some images seem to have a weird discontinuous histogram with separated sticks like that (cf file attached) and some have a classical continuous histogram...)

Has it ever happened to anyone?

LL

[Alexandre Dufour]

Hi LL,

This could also be due to a special option within the acquisition software that "stretches" the histogram. That may happen when you adjust the contrast before imaging, and the software wrongly applies this contrast transformation directly to the acquired image (I've seen it happen a lot). Could that be the case?

Alexandre

--
You received this message because you are subscribed to the Google Groups "Icy imaging" group.
To unsubscribe from this group and stop receiving emails from it, send an email to icy-software...@googlegroups.com.
To post to this group, send email to icy-so...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/icy-software/f61ec24a-bd12-40d8-acb5-03a334806359%40googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
<Discontinuous Histogram.docx>

[louise-laure]

Hi Alex,
sorry for the delay, I'm sure I'm writing the middle of the rush post new year break!
but I wanted to check a few things before answering back.

(We also had to change our detectors because one broke and I tought maybe the problem with the histogram was linked).
After a few investigations, bottom line is:
the weird histogram is not due to the softwre options, but specific to these detectors per se.
There seems to be an incompatibility with the Leica software and Hamamatsu detectors,
even though the information of various levels of grey is there
the correct histogram is not retrieved and shown... I don't why yet

so if anyone has an intel about this!
(we checked with our older detectors and the histogram is normal),

So I'm back to defining manually the contours of my ROIs...
which is pretty annoying and oldschool!

If you have any way to help me with these image,
or need some new images (maybe you threw away the old ones), please let me know!

Thanks in advance,
LL

[louise-laure]

And I may also add:
These issues with the histogram were also present in my images in which we already managed to find the cells in 3D.
(cf with Alexandre's protocol (cf post: "How to measure volume and suface of a cell filmed in a .tiff 4-d stack)
to set the parameters of 3D active mesh plugin to find my cells and follow them over time)

so I guess the histogram shouldn't be an issue to find the cells,

so For dye n°2, (Ca imaging)
since some of my cells only seem to appear at some precise time point when they fire an Action potential,
I'm back to asking you if it's possible to find the cells at a specfic time point and then keep those automatically detected ROIs at the same location over the all time frames.
If necessary/better, this could be done in 2D only since my stacks are only 3 to 4 z slices thick

Thanks +++,
LL

[falcok...@gmail.com]

Hello together,

as I was searching the web for any publications and hints for what is the reason for these weird HyD histograms and if maybe somebody used these for data analysis before, I found this group.
I wanted to perform colocalization analysis of data sets acquired at our Leica SP5II with two Hybrid detectors. Everything was fine until I saw the Coloc2 and PSC scatterplots. Their appeared like a grid, not continuous as one alsways sees them in publications.
I checked everything I could think of, BrightR mode or standard didn't give different results. Also the normal pixel histogram of the images look like the already mentioned spikes.

Before using smoothed/blurred data for analysis and publish that, I contacted a Leica technician yesterday... (This is a translated summary)

As it turns out, it is not a problem of the HyD, it is the advantage of the HyD.
He told me that the distinct histograms and scatter plots arise from the very precise
spatial as well as temporal photon acquisition resolution of the HyDs. This is true for
both, the Standard as well as BrightR mode. Meaning, that the "usual" histograms with
their continuous pixel intensity values arise from a blur effect due to their unprecise
acquisition.
Furthermore, he assured me that we can/should use this data without post-processing even
though the scatterplot may look weird, all pixels (their intensity and number) are
included and contribute to the plot and the calculation. It is only a mistake in beauty,
not in analysis or acquisition.
So the only problem left is only that everyone is used to the smooth scatter plots.
Roughly speaking, the smoothness is indicates unprecise photon acquisition.

As this may not solve your problems, you maybe know now why the histograms look like they look.

Best,

Falco