Help with segmentation of clumped cells

fiji
imagej

#1

Res_1|nullxnull Lyso_1|nullxnull
I’m having trouble getting segmentation of the cells in the attached image.One of them is from a raman spectroscope (so looks more like a phase contrast image and one is a normal fluorescence image of lysotracker). I included both as the segmentation can be done on either.

I’ve tried a number of things and not really achieved much but I’ve seen what I thought were hopeless cases on this forum get solved so hoping someone can work some magic or has any specific tips as I’m feeling a little lost.


#2

Good day,

your task is by far the worst case I’ve seen for long …

Although I’m not trained in analysing such images, I feel unable to do a manual segmentation (in both images) guided by eye. In such cases there is no real chance to get reasonable results from a machine.

Regards

Herbie


#3



#4

These are the identical images you’ve posted before — no?

Any further comments?

Regards

Herbie


#5

Hello segmentation Help,
Which part of the image are you trying to segment? The bright parts? The ring shaped structures, or both? Just asking for clarity.
Bob


#6

Yes sorry, I believed the upload hadn’t worked so I uploaded them in a different way but my account got temporarily suspended so the upload was delayed.

Thank you for your opinion. Makes me concerned about what I’ve been seeing with my eyes!


#7

Hi Robert,

Ideally, I would be able to collect lysosomes per cell so my image below shows what I THINK is the segmentation of cells.

Another option, if possible, would be to segment the nuclei and just spread out a certain distance from them and count the lysosomes in there. I thought this might be possible since in some of the cells there is a nice contrast around the nuclei but obviously it’s not there in all of them.

Not feeling that hopeful though!


#8

Makes me concerned about what I’ve been seeing with my eyes!

Sorry but I don’t understand this, if it was meant in response to my reply.

The thing is that if you are able to perform a task visually guided, it doesn’t mean that a machine can do it. I’m pretty sure that this holds for your sample images.

Herbie


#9

What I meant was, since you say

I feel unable to do a manual segmentation (in both images) guided by eye

That I am just convincing myself that I see the demarcation between cells that isn’t there.


#10

As I’ve written with respect to your sample images: “I’m not trained in analysing such images”

I didn’t write that you see something that isn’t there!

A trained medical X-ray specialist will see details in an X-ray image that most people won’t realize … That said, you can get trained to evaluate a certain kind of images.

The human visual system is extremely good, e.g. in finding contours but not, e.g. in making statements about absolute properties such as gray values etc. It isn’t suited to measure. For the latter, machines are far better…

Regards

Herbie


#11

Hello Help (using the short name),
Don’t get disappointed so soon. You never know what may happen.
Bob


#12

Not that familiar with a Raman spectroscope, but aren’t you able to tune the bands to peaks for DNA and RNA to get “images” with the nucleus only; and other bands to perhaps get the cytoplasm??

If not, you can start with using Hoechst as a live nuclear marker…
What cells are these?