Hi Mike,
Have a look at the binary image here
https://dl.dropboxusercontent.com/u/11163658/TEST.jpg-%28Colour_1%29%20Maxima.tifI used the test image you uploaded and did the following. A couple of notes, firstly I am using the Fiji distribution (www.fiji.sc) of imagej so you may not have the plugins or filters I used. Secondly the image is a very low res, the nuclei only have a couple of pixels in them so it makes finer detailed selection hard. But give this a go and see what you think.
1. Colour deonconvolve the image. Image --> Colour --> Colour Deconvolution. Set the vectors to H and E. you will get three images. Colour 1 is the haemotoxylin channel, colour two is the eosin and colour 3 is essentialy a mathematical remainder. If you look at colour 2 you can see the clumps already, you could threshold this and just measure that but it won't give cell number etc.
2. Select the Colour 1 image, invert it (Edit --> Invert) and set the LUT to gray. This is purely for aesthetics, I just find it easier to work on "fluorescent type images".
3. Flatten background (Process --> Subtract Background) using a 50px rolling ball and all other options disabled
4. Detect the individual nuclei with the find maxima command (Process --> Find Maxima), set the noise tolerance to 15 and set the output type to single points
5. Duplicate the single points output for later.
6. select one of the single point images and invert it (edit --> Invert) now apply the distance map to this inverted image (process --> binary --> distance map). Now invert this result (edit --> invert).
7. The resulting image won't look great, you can adjust the brightness and contrast a bit but it will still look mostly saturated. But it doesn't matter it will do the job it's supposed to.
8. Set a threshold on the final distance map image of 253-255 and apply it to get a binary image.
9. Now run an Open filter on the binary image by going to Process-->Binary-->options. Set iterations to 2 and count to 1, tick black background and chose do to be open.
10. you now have a mask to use for further analysis. I had a quick play and did the following.
11. Use the analyse particles option to exclude the smaller single clumps (size 30-infinity). Out put the results to the ROI manager (tick the add to manager box).
12. now select the other single points image you duplicated earlier. Select the roi manager and tick the show all box to make the regions appear on it. Now set your measurementsanal (analyse --> set measurments) to measure just integrated intensity. Now click the more>> button on the ROI manager and select multi measure.
13 the results will be a totally/integrated intensity for each ROI. As the image measured was single pixels of 255 intenisty, dividing the result by 255 will tell you how many points/cells per cluster.
Cheers
Cam
Cameron J. Nowell
Centre for Dynamic Imaging
The Walter and Eliza Hall Institute of Medical Research
1G Royal Parade
Parkville, Victoria 3052
Australia
Phone: +61 3 9345 2871
Mobile: +61422882700
Fax: +61 3 9347 0852
Facility Website
LinkedIn Profile
-----Original Message-----
From: ImageJ Interest Group [mailto:
[hidden email]] On Behalf Of Machado
Sent: Tuesday, 14 May 2013 2:40 AM
To:
[hidden email]
Subject: Re: Outline Area of Specific Cell Density
Hi Cam,
This sounds like a great idea, but I can't get the distance map to recognize the correct cell clusters. In the images below, you can see the distance map that I created, and the thresholded image. It seems to be picking up on some noise near the middle of the tissue section. In the third image, I've just changed the type to 8-bit, and then thresholded to determine the cell clusters as I see them (Default, B&W, 0, 120). Of course, converting this to a binary image removes this threshold. Is there a way to use the 8-bit thresholded image that I've created and still create a distance map of sorts? Or perhaps use the grid to output number of cells per grid location?
My thinking is that I could write something into the macro at this point to recognize my "cell cluster" specification (e.g. >10 cells/grid), and then either highlight these portions of the grid, or fill in the other grid squares in order to get an area value for these cluster-positive locations.
Thanks!
Mike
<
http://imagej.1557.x6.nabble.com/file/n5002981/TEST1.jpg>
<
http://imagej.1557.x6.nabble.com/file/n5002981/TEST2.jpg>
<
http://imagej.1557.x6.nabble.com/file/n5002981/TEST3.jpg>
Mike Machado
Research Associate
J. David Gladstone Institutes
Institute of Neurological Disease
San Francisco, CA
--
View this message in context:
http://imagej.1557.x6.nabble.com/Outline-Area-of-Specific-Cell-Density-tp5002970p5002981.htmlSent from the ImageJ mailing list archive at Nabble.com.
--
ImageJ mailing list:
http://imagej.nih.gov/ij/list.html______________________________________________________________________
The information in this email is confidential and intended solely for the addressee.
You must not disclose, forward, print or use it without the permission of the sender.
______________________________________________________________________
--
ImageJ mailing list:
http://imagej.nih.gov/ij/list.html
Locked
