Automate the ROI identification and selection in fluorescent microscopy images

Hi everyone, first of all I would like to wish you a happy Christmas.

For my graduation project needs, I am trying to write a python script in order automate the identification and the selection of the region of interest in a fluorescent microscopy image.

The identification of a region of interest can be done following several criteria, and that is what I am trying to identify first, as I am not a biologist or an expert microscopist I have no idea about the more used methods of criteria to track a region of interest.

I created this topic, in order to collect tips and advices from experimented fluorescent microscopy user or scientist that used to identify a region of interrest in a sample.

If anyone have any information that can help me, such as idea of criteria to track ( size, brightness intensity, shape, density of object…) Or anything else (papers, library,Codes…) I would be very grateful if you let me know…

Thank you !!

Good day,

without seeing any typical images we can’t really help you. There are hundreds of different fluorescent stainings and thousands of different tissues and each requires a special way of detection/segmentation etc.

Did you already search the Forum to get an impression of the large variety …

Good luck


Hi Herbie, thank you for your fast answer.

Yes I did, I already took a look at several papers, forums etc …

The main reason that made me creating this topic, is to collect tips from experimented scientist in the field, to help me to choose at least some of the most used methods, or some of the most pertinent criteria…( i am really sorry for being not precise and that confused… i do not want to limit anyone who can help me, i am open to all proposition)

I am well aware that at this point, we can’t automate the selection of ROI in all microscopy imaging…
but we can start by automate the ROI selection in some of FM images .

as I’ve written and substantiated:

without seeing any typical images we can’t really help you.



I think that you are not understanding me, i am not aiming to detect regions of interest in a specific image, i am trying to automate this detection in different kind of images… not a special ones…

Thank you for trying.

As I’ve tried to point out before, there is no universal method.



As you may guess, I am well aware about that and I tried to point it out before as well. thank you for trying to help me.

just do a course on image processing and computer vision.
it’s full of them nowadays on the web.

Hi emartini, thank you for your reply.

I actually have some skills in image processing, but a lack of knowledge in biology and microscopy.

what i would like to know is how can we define a region of interest, not a specific one, but in general. i am aware that there is no magical recipe, and that my question is not that evident… but i may find some answer to this kind of questions.

what kind of question could we ask to the user in order to help him to find the region of interest… what kind of criteria are the most tracked, (size, shape, density…)…

and i am open to all kind of proposition, ideas, papers…etc

what kind of question could we ask to the user in order to help him to find the region of interest

The user knows to locate the ROIs (Regions of Interest) because the user is just interested in certain regions of images. However, the user may need help concerning selecting ROIs by suitable software tools.

The problem is that nearly every user is interested in regions of images that are differently defined.



1 Like

dear Herbie, Hi again, i am sure that every single image have it’s own region of interest, and every fluorescent imaging process aim to figure out some thing special.

is there any kind of specifications, that we can track to facilitate the user work. for sure every user is interested by a specific criteria, but is there any common specifications that we can automatically locate in the sample, that the user will directly search there and not in the whole sample ??

The answer is simply no.

Your project appears slightly ill-posed to me.