I will have a set of fifty 32bit, ~2-10GB pyramidal whole slide images in ome.tiff format that I would like to convert to 8bit pyramidal whole slide images (including multiplying the original 32 bit values by 100). Does anyone have a software recommendation to accomplish this?
Each image (unfortunately cannot be shared, but I can probably mock up something from several open source images) was written out using QuPath from a set of 8 deconvolved images. The final images have 9 channels and are generally around 7GB. The default bit depth there is 32, as it handles negative numbers and all of the decimal places needed to cover the optical densities from the deconvolved channels. I checked, and there does not seem to currently be a way to export the image “with some math.”
Pyramidal ome.tiff, 32bit, 9channel, values ~0-2.5 -> ?? 8bit, 9channel, values 0-255 (truncated on either end)
All values in all channels tend to be in the range of 0.01 to 2.00. There are some values below 0 and above 2.55, but I am fine with clipping those as they are not generally biologically relevant (black smudges, or below significant thresholds). So my hope was to multiply by 100 and then save as a 0-255 8bit image (per channel). And hopefully this will reduce the file size a bit.
My first thought was FIJI. The one image I have tested did open in FIJI (after downsampling they are only 22,000 by 25,000 or so pixels), but it took a very, very long time to open. I am not sure how long as I ended up going to bed and it was finished in the morning. So a macro there might be the only way to go if I can get everything written out well. The image math is easy, but I am not sure how to convert to 8bit while clipping values above and below 0 and 255. I think the default 8bit command interpolates based on max and min values, which would be bad for consistency.
Does anyone have any other suggestions? I looked at the bfconvert commands but was not able to find anything about performing pixel math or changing bit depth. I am also considered Imaris, but the image math there would be manual per image, and the output would be a .ims file (which would be an additional step to convert back to something like a pyramidal ome.tiff).
Anyway, I will cut off my rambling there, but please ask if you have any more questions or I can specify anything else useful about my inputs or desired outputs.
*None of this is for quantification, as all measurements were made from the brightfield images. I am not extremely worried about the exact accuracy of the pixel values as long as they are generally correct. This is purely for visualization of the different stains in combination with cluster analysis/gating.