Counting nucleus on tiled stacks of images

Hi all,

I am imaging mammalian embryonic jaws at the biphoton microscope (SP8 DIVE from Leica). To be able to image this jaws, I tile the entire jaw (between 8 and 20 tiles depending on the developmental stage), each tile being a stack of file (500 to 1500 µm thick ~300 image per stack). My files are between 6 and 20 Go.

These jaws have been treated with EdU to detect cell division, so all the nucleus of dividing cells are labelled (with a 647 dye). I now want to see where the cells division occur and the local density of them within the jaws. My plan was to:

  • visualize the jaws in 3D to locate the cell division within the jaw
  • count the nucleus using 3D OC

Here are the issues that I have:
*despite having a powerful workstation (xeon gold proc and 128 Go of RAM), I sometimes have issue with the 3D viewer. Is there a way to allocate more memory? Would an ImageJ based software like Icy would be better for 3D visualization?
*I have trouble running 3D OC because the images are obviously big. Is there a way to reduce the file size or to make it better handled? Is there another plugin that would be better for this type of task?

Thanks a lot for your help!

Please forgive me if I am just being ignorant. Isn’t 3D Viewer part of ImageJ? For sure, I know ImageJ has a plugin that goes by that name. If you are using the ImageJ plugin, then you might want to check Edit > Options > Memory & Threads....

We just did analysis/visualization in ImageJ via SciView for EdU markers in mouse incisor with similar file sizes.

You might simply have issues with the size of your image data with respect to the amount of graphics card memory. We are currently testing “out of core” volume rendering in SciView via some new improvements based on BigDataViewer, which serves as a workaround for images that exceed graphics card memory.

I suspect you’ll run into some other issues, like the need for better segmentations of the nuclei. That said, when you do analysis you might want to analyze them in tiles, which can make 3D OC tractable. But also check out what @haesleinhuepf’s recent work on CLIJ, I suspect he knows how to solve the nuclei counting problem.

@Andrew_Shum Yes, sorry for the confusion, I am using the 3D viewer plugin in ImageJ but it is very slow (because the images are big) even while putting the memory in max sadly.

@kephale awesome, I was also thinking about doing something “out of core” based but I unfortunately don’t have a lot of good IT support in my department here at UCLA. If you come up with a solution that works and can share it I would love to hear how you implemented this. I’ll check this work on CLIJ as well! Thanks all for your help!