Z-axis intensity correction

What is the best method to correct for loss of signal intensity in the Z-axis? I am having trouble with image segmentation/ pixel classification (using ilastik or with thresholds) due to intensity decay through the Z-stack. While I do not need to quantify intensity itself in the images, I do need to quantify the number, area and shape of classified puncta and Z-scatter/aberration is interfering with classification due to varying intensity. I came across this publication and I was wondering if anyone has tried or would recommend the Intensify3D program: https://www.nature.com/articles/s41598-018-22489-1

This post seems relevant: Compensating brightness attenuation in a confocal stack?

Is there any consensus/optimal method for this sort of correction (in any program, doesn’t need to be program-specific)?

Thanks!

1 Like

Talley pretty well covered it in his first post. It is a difficult problem, and when information is lost during sample imaging, it cannot “really” be recovered, just estimated.

That is part of why large object light sheet imaging involves rotating the sample .

1 Like

Hi there,

I’m working on a similar project as you’ve described and also wanted a method to account for decreasing intensity along Z. I looked into Intensify3D, and while the results seemed promising, I wasn’t able to get the program to run either on Linux or Windows.

This led me to implement the algorithm myself in Python, which you can feel free to check out here: https://github.com/dakota-hawkins/intensipy . There are some differences between the original implementation – at least as best as I could discern the original source code + publication – but differences are documented and original implementations are optionally available.

There’s some cursory results you can look at on the github, but we’ve generally found it very useful in our work. Of course, as you mentioned, with this method you’ll want to stay away from inferring too much meaning out of the intensity values themselves.

Best!

1 Like

Hi @Paulahyns,

This macro could be worth to check out:

1 Like