Correct Illumination with Higher magnification / larger objects?




I’m running into a bit of trouble trying to perform illumination correction with images of mammalian cells under higher magnification (60X). The images were generated using an ImageXpress XLS microscope, and the size of each tiff image is 2160x2160 px. Since most of the image is background (staining nuclei and a viral particles), I opted for the “Background” mode of CorrectIlluminationCalculate. The problem is that the DAPI-stained nuclei are ~100-200 pixels in size, meaning the block size needs to be at least 200 - 250 pixels to ensure that each block contains background pixels. This leads to very blocky correction images.

So, I have the following questions with respect to how to proceed:

  1. Should I simply use a smaller block size and use “all cycles” to generate an illumination average for all the images, or a larger block size and use “each” / “all cycles” ?

  2. How should I balance the block size and the median filter size for smoothing? If I have a block size of 200px, what would be a reasonable filter size for smoothing?

  3. For other channels with smaller objects, is it reasonable to use a smaller block size? Or should I generate an illumination correction image for each channel using the same settings?




Creating an illumination correction function across all the images is definitely one option! It will allow you to normalize intensities across the field of view. If you’re just trying to remove background to improve segmentation though, that may not be what you’re looking for.

We generally manually adjust the smoothing size, so unfortunately I don’t have hard-and-fast guidance.

We typically use the same settings for all channels, but I haven’t rigorously tested whether for objects of very different sizes which would be the better approach- if you find one is better vs another, I’m curious to know that answer!