Analysis of fluorescence images

[GABRIELLY RIBEIRO]

Hello, recently I started working with QuPath and analyzing light microscopic tissue images is quite impressing in terms of segmentation and classification.

Lately however, I have been trying to segment (identify) nuclei of an image with three fluorescence channels: CD14, CD45 and DAPI. The images were generated using a laser scanning cytometer. 

Since it does not seem possible to load several images at a time, the three images were merged and saved as color image (RGB). But the identification of objects of interest (i.e. cells) stops at the TMA. The best result so far is attached. May anybody tell me how to make it better? 

Any help would be greatly appreciated.

[micros...@gmail.com]

I see a couple of things in your picture I have questions about.

1. Both your picture and your description seem to indicate that this is not a TMA.  Why are you starting with using the TMA dearrayer?

2. The image you loaded may be problematic as well, as both the "ImageJ server" and the lack of any pixel information at the bottom of the left window indicate problems reading the file as you input it. 

You may just want to do a simple tissue detection with the default settings to get the entire square (should happen), and then attempt a basic cell detection on your DAPI channel using the entire area.  Your measurements will all be in pixels, and there are a few other commands that will not work well without pixel size information, but it should get you started.

Also, I cannot see from the image, but the Image Type is set to Fluorescence, correct?  Not having it set (which can happen with altered images) could cause problems as well.

[GABRIELLY RIBEIRO]

Thank you for considering! 1. I could not keep up with all the updates, but I started with the TMA dearrayer because without it I can not identify the cells. What would the other TMA suggest? 2. The pixel size is shown in the image on the left side. The image type is set to fluorescence. Could you give me some help figuring out the exact settings to make better progress? I need identification of all objects.

Thanking you for your help!!!

[micros...@gmail.com]

It doesn't look like you have a Tissue MicroArray (set of circular tissue biopsies on a slide), so I would recommend either hand drawing the area (https://github.com/qupath/qupath/wiki/Drawing-regions), or usually with fluorescent images the Simple tissue detection command will result in a box that selects the entire slide.  Since you don't have any pixel information (pixel height and width are Unknown), the settings for cell detection will be a bit... odd, so you may have to play with the values a bit.  

BMP images also do not usually have color channels, so I am not sure how well the cell detection will actually work, but if you are not having problems with it, great! If your image has separate color channels, or good enough separation, you should be able to perform Cell detection on any hand drawn or simple tissue annotation.  Just find the DAPI channel (visualized most easily by pressing 1,2,3, or 4) and you should be able to follow a similar workflow to the wiki (https://github.com/qupath/qupath/wiki/Detecting-objects).

While I have done sample cell by cell analyses off of a PDF file, so this should be doable, you are probably better off combining the images in another way for fluorescent analyses.  TIFFs usually are split into color channels that make analysis easier.  I do not know what your setup is for taking the images, though, so I cannot really help much there.