virtualhe_rust

This tool creates virtual H&E images from fluorescent microscopy data

The implementation is is written in rust, is NOT GPU accelerated, and is based on the following paper:

Giacomelli MG, Husvogt L, Vardeh H, Faulkner-Jones BE, Hornegger J, Connolly JL, Fujimoto JG. Virtual Hematoxylin and Eosin Transillumination Microscopy Using Epi-Fluorescence Imaging. PLoS One. 2016 Aug 8;11(8):e0159337. doi: 10.1371/journal.pone.0159337. PMID: 27500636; PMCID: PMC4976978.

Description:

• Precompiled binaries are available for:

  • Windows: bin/win/virtualhe.exe
  • Linux: bin/linux/virtualhe

  A Python implementation can be found here.

This tool takes a 16bit or 8bit greyscale image of nuclei (vHematoxylin) and a 16bit or 8bit greyscale background image (vEosin) like autofluorescence or a fluorescent counterstain like eosin. The output in a 8bit RGB image saved to disk.

Usage example:

bin/win/virtualhe.exe --help

--- BEGIN OUTPUT ---

Make a Virtual H&E Image from Fluorescent Microscopy Images

Usage: virtualhe.exe [OPTIONS] <NUCLEUS> <EOSIN> <OUTPUT>

Arguments:
  <NUCLEUS>  Path to the nucleus (hematoxylin) channel image (e.g., nucleus.tif)
  <EOSIN>    Path to the eosin channel image (e.g., autof.tif)
  <OUTPUT>   Path to save the output RGB image (e.g., output.tiff)

Options:
  -k <K>      K arbitrary factor to adjust color profile of H&E [default: 2.5]
  -h, --help  Print help

--- END OUTPUT ---

• Now we will take 2 images:
  • Nuclear Image (i.e. DAPI, sytox, topro3): c:\data\images\nucleus_image.tif
  • Background Image (i.e. autofluorescence, eosin): c:\data\images\vhe_output.tif
• We want the output image to be located here:
  • c:\data\images\vhe_output.tif

# Run this command
python c:\virtualhe\bin\win\virtualhe.exe c:\data\images\nucleus_image.tif c:\data\images\background_image.tif c:\data\images\vhe_output.tif

--- BEGIN OUTPUT ---
Reading c:\data\images\nucleus_image.tif
Reading c:\data\images\background_image.tif
Scaling channels
Calculating and Saving vH&E
Virtual H&E image saved to: c:\data\images\vhe_output.tif
--- END OUTPUT ---

NOTE:

• Memory: The amount of memory required to process is approximately 10x the size on any 1 uncompressed 16bit input image and 20x the size on any 1 uncompressed 8bit image.
  • For the example above, if the 16bit image "nucleus_image.tif" is 1 gigabyte, it will require ~10 gigabytes of RAM to process (1gigabyte * 10).
  • If the image was 8bit, it would require ~20 gigabytes of RAM to process (1gigabyte * 20).