Analysis of Prussian Blue

Hi All,

I’d like to do analysis of both the intensity and the amount of Prussian Blue stain visible within a tissue section.
Could you please help me with that?
Thank you!

Welcome to the forum, @Daria!

Have a look at the Colour Deconvolution plugin (Image > Color > Colour Deconvolution in Fiji). If you need more specific help, please upload a sample image.

1 Like

Hey @Daria - Welcome to the Forum!

We would love to help you. To get started - would you be able to post an original image here and give us a more detailed description of what exactly you wish to measure?

Many thanks!

eta


Hi @etadobson Thank you! Please see the attached image. I’d like to measure the area covered by the blue stain as well as the intensity.

@imagejan Hi, Thank you - I uploaded a test image. I hope that helps a bit. Could you share the step by step protocol for the analysis (area and intensity of the blue)?

Hello @Daria,
You first convert your images to an 8 bit format and then use the ‘Threshold’ tool (under Image>>Adjust) to ‘highlight’ the blue spots. Try not to click the ‘Apply’ button in the Threshold tool, as this would create a binary image (0-255). Once you have done this, go to ‘Analyze Particles’ (under the Analyze tab) to measure the area and intensities of all your thresholded regions. You should then have your results displayed in a table. Hope this helps.

-Praveen

@Daria

Thanks for posting your image… That helps the community answer your questions more clearly.

When it comes to quantifying images - it is best to store them as tiff files - as opposed to jpeg - in order to avoid lossy compression issues. Read this over - it explains why jpeg storage can result in loss of information.

So if I were you - I’d start acquiring your datasets as tiffs (and making sure your acquisition parameters match for each image). But then yes - you can follow @Praveen’s and @imagejan’s (Colour Deconvolution might be the best place to start) suggestions for the analysis of your blue-stained cells.

Best of luck! And post again if you have any more questions in the future for this or any other analysis issues.

eta :slight_smile:

Hi Praveen,
I followed your instructions and I got over 7000 readouts in my Results table. When I went with the “Measure” (under Analyze>>Measure) I got a specific value, however when I run the analysis for an image with no blue stain, the difference was very minimal. It appears that something is still not right when it comes to assessment of blue color.

@etadobson
I followed Praveen’s advice, however there still appears to be some issue with the analysis of blue color, since image with blue stain and one without it produce very similar values.
I am not familiar with the specific protocol for the Color Deconvolution. If you could share the step by step protocol, that would help.
Thank you!

Hi Daria,
Thats strange. I just checked your imagea again and got around 2300 spots, and it seems to be running fine. It also gives my the mean intensites for each ROI selected. Here is a screen shot.

Are you sure you converted it into an 8 bit image? Could you also maybe upload a screenshot of your thresholded image?

-Praveen

Hi Praveen, You’re right - when I run the analysis measuring number of particles I do get similar read-out, but what I am mostly interested in is the area covered by the blue and the intensity of the blue. Both the intensity and the area are important to distinguish between different treatment groups.
Would such analysis be possible with this image?

Hi Daria,
Yes this would be possible. After running Analyse Particles, the results table should have the mean intensity for each ROI (under Mean). The table should also show you the area of that ROI as well (under Area).

-Praveen

Hi Praveen, the protocol works well for the first image where the particles are clearly distinguished from one another, however it still doesn’t seem applicable for the remaining images. I’m including a representative picture of the group of images that I need to quantify (there are several groups). When I look at Area and Mean, the values don’t make sense as it seems that the thresholding is not ideal and it picks up other particles/shades as well, instead of only blue.
Thank you for helping me with this,
Daria

The two images you provide in the first and the last post apparently differ in their white balance as well as in their magnification and exposure. You can’t expect a method developed on the first to work robustly on the latter one as well.

First, you should make sure you use exactly the same acquisition settings (i.e. no auto-exposure, no automatic white balance) for all samples you want to compare/analyze. Only after getting the most consistent acquisition possible, it makes sense to invest time into a robust analysis.

That said, if you still see high variability in you images, try using machine learning (Trainable Weka Segmentation or ilastik) on a truly representative subset of your images to train a classifier for segmentation. Also have a look at these discussion:

4 Likes

@Daria
I would have to agree with @imagejan here. It would be hard to follow a single method to do the segmentation if the pictures differ a lot… Have a look and see if you can acquire images with the same settings while you capture them.
We are here to help as always :slight_smile:

-Praveen

1 Like

Hi Praveen, could you please help me with the protocol that would assess the intensity and area of the second image that I sent you? The pictures differ because the tissues are different and we are looking at different structures within the tissues, but they are not meant to be compared one vs the other. The Image 2 is an example from a large pool of images for this specific group. Let’s focus on the Image 2. I’d appreciate your help if you could list the steps that I should take to arrive at the measurement results.
Thank you :slight_smile:

Ok,I can try. But could you maybe upload some high resolution pictures (like a Tiff format). The one which you show here i is quite pixelated which makes it more difficult to work with.

-Praveen

Hi Praveen, sorry for the delay in communication.
I just took a couple of TIFF images, however the site is not letting me upload TIFF, they only allow for jpeg, png, zip, gif. so I saved them as jpeg. Hope these will work.
Thank you,
Daria

Hello Daria,
You can also compress your TIFF in a zip file and then upload it. I tried with the new JPEG files you uploaded (which is only 800x000 pixels), but it is not perfect. If you can provide us with TIFF files it would be better i suppose.

Nevertheless here is a screen shot of the segmentation of one of your images done with the Trainable Weka Segmentation plugin. Have and look and see if this you are happy with the segmentation. If yes, i can continue and tell you how to do the rest of the analysis.

2 Likes

Hello Praveen,
The image you sent looks great! When I focus on the purple areas only, it seems to be exactly the areas of interest.
This is wonderful! Please do share step by step instruction how to do the analysis. :slight_smile:
Thank you!!