Analyzing adjacent ROIs

[Alexander C. Taylor]

Hey folks, I'm new here and I'd appreciate some help if anyone has a few minutes. I hope I'm posting this in the correct area.

I have some long bone sections with staining, and I'm trying to find out if there's a gradient, as in more staining towards the center compared with toward the periphery.

From what I can find, there seems to be no option to give ImageJ a rectangular ROI and ask it to compute staining gradients, so I've been trying to find a way to align multiple small rectangles: [][][][][][][][] from the outer edge of the bone in the image to the center of the bone. These rectangles would need to be uniform and perfectly aligned, and I'd analyze each for total staining to see if this increased in a given direction. These rectangles would be at an angle, since I have to go from a certain spot on the outer edge of the bone sections.

Is there any way to achieve this, or another way to determine something like this? I'm at a loss.

Thanks, folks. -Alex

[Michael Doube]

> I have some long bone sections with staining, and I'm trying to find
> out if there's a gradient, as in more staining towards the center
> compared with toward the periphery.

A bone question! Excellent.  Are your sections roughly annular, like
transverse sections through the cortex of a long bone?  Could you post
an image of your section on the web somewhere we could see it?

I can imagine one solution to this; draw an ROI that is the periosteal
surface and another ROI that is the endosteal surface. Then (hard part,
don't know if it's implemented) create multiple new ROIs interpolating
between the landmark ROIs, and measure the staining between adjacent
pairs of ROIs.

I just thought of a way to do the interpolated ROIs, so it can't be
*that* difficult...

Michael

[Alexander C. Taylor]

Thanks for your reply, Michael! Yes, the sections are fairly annular- and yes, they're sections of long bone. Here's a link to an image similar to the ones I'm dealing with: http://decembermouse.webuda.com/Sample/sample.jpg

As long as I can create ROI's of equal size and have them perfectly adjacent and aligned, we should be able to find what we're looking for... I'm relatively new to ImageJ but think I have a handle on all the basic functions.

I appreciate the help!
Alex

[Michael Doube]

Hi Alex

> Thanks for your reply, Michael! Yes, the sections are fairly annular-
> and yes, they're sections of long bone. Here's a link to an image
> similar to the ones I'm dealing with:
> http://decembermouse.webuda.com/Sample/sample.jpg

Thanks for the sample image.

A more immediate issue appears to be how you plan to measure 'staining'
at all, even before attempting to show any spatial trend, and what that
measurement might seek to represent.  There are quite a few problems
with measuring the strength or amount of histological staining, which
get discussed on this list from time to time.

The morphology of the section shows shrinkage and sectioning artefact,
uneven staining and tearing, so you'd have to work around that as well
(some dark stripes look like they are due to knife chatter for example).
 It looks like the most prominent staining is the edge of the marrow
where it has come away from the periosteal surface.

So while it's possible to proceed with some kind of spatial analysis,
I'm not convinced that your image actually holds useful information of
the kind you are seeking.

Michael

[Michael Doube]

> where it has come away from the periosteal surface.
*Endosteal*, of course.

[Alexander C. Taylor]

It's not our best image, just an example of the type of tissue we're looking at. I agree that the morphology shown here could pose some issues, and we're still looking at ways to remedy those or find ways of analysis that those issues won't hinder. There are some knife marks and shrinkage, and the staining is uneven on some of the early slides. This was a section we used to develop our staining procedure, but some of the final images do have similar, albeit less dramatic (fortunately), detriments.

I'm not too familiar with general problems affecting measurement of histological staining; I'll look around for posts on those to learn more. Thanks for the tip.

Fortunately we're not looking at cortical bone, just at the marrow. I probably should have mentioned that in my last post, come to think of it. I'm just trying to get a not-terribly-sophisticated measurement of whether staining intensity has some sort of gradient from the middle of the marrow to the edges or vice versa, which if it exists, will affect the direction of our study. Using ImageJ, I have been able to Set Threshold so that only areas of staining are highlighted, so there's that.

I apologize for not posting one of the higher quality images; my P.I. wouldn't condone posting those online. But as long as I can get some idea on how to work with this type of tissue section and look for a gradient, we should be able to find what we're looking for.

[Michael Doube]

> I'm just trying to get a not-terribly-sophisticated measurement
> of whether staining intensity has some sort of gradient from the
> middle of the marrow to the edges or vice versa,
A really simple way to do this is simply to draw a line across the
marrow, (double click line tool icon to change line width), then hit [k]
on your keyboard to get a profile plot of pixel value vs. distance along
line.  A gradient will show up as a slope in the line. This will of
course also include the trabeculae if you leave the images as they are;
you might have to do some segmentation before a profile plot gives you
results on marrow only.

Michael