Object cell area tracing


#1

Hey all,

With the help of the forum, I’ve solved how to measure separate RGB intensities using the same selection. My next question relates to making an actual selection. The purpose of what I am doing is, using skeletal muscle stained for three differing myosin heavy chain isoforms, to quantify fiber type and area based on the relative intensities of each isoform (e.g. if a fiber has 75% red, 20% green, and 5% blue- I would quantify that fiber as a predominantly ‘red’ phenotype).

My ‘issue’ is the lack of consistency I am able to produce by manual drawing of the fibers. I have played around with threshold function a bit. However, it does not seem to truly ‘get the job done’.

Do you guys have any advice as to how I can more reliably make the area drawing, then use the area drawing to measure size and relative intensities?

Thank you. I am attaching a sample image for reference. An accessory problem is the lack of membrane stain for our samples. However, we are unable to run the samples again.


#2

@Kev

Would you be able to show us what constitutes a ‘fiber’ - ie. the object(s) you want to segment? Just annotate what area you want to segment…

Honestly - I am not familiar with your type of sample… so I am not quite sure what I’m looking at.

That’s the first step in getting you some help.

eta :slight_smile:


#3

Yes of course etarena! No problem.

So each (quasi) circular object is a cell. There are probably 100+ cells in this image. They are different colors based on the cell type (muscle fiber type). For example, you can see several red stained objects, those are a specific cell type- just as predominantly green or blue objects within the image are other cell types.

Basically what I am doing is tracing around each single object, regardless of color, then performing essentially an integrated intensity analysis for each wavelength of color (RGB).

Does this help? If not, I can trace some fibers in Paint and re-upload to better display what I am doing.

Thank you for the help!


#4

Your description is clear @Kev - Thanks!

… for getting started in Segmentation in ImageJ/Fiji - here are a few helpful links/tools you can look into to start:

I will take a little time to try some things out… but you can look into the above links/tools in the meantime.

eta :slight_smile:


#5

@Kev

I did a quick test with Trainable Weka … and this was the Segmentation result I got with just the default settings:

It did a good job I’d say… and could be better creating a decent mask you can then use as a base for further analysis with a tool such as MorphoLibJ.

eta


#6

Thank you very much for your helpful suggestions! Sorry for the delay on response, I’ve been out of town for a conference and only have just returned. I will take a look at all of these items and get back to you on here. Again, I appreciate the help tremendously!