Color palette for annotating histological images

Todd Valerius 2020-06-02

Annotating histological images requires drawing outlines of structures on the image and adding a Class name (the anatomical term from the approved ontology). Typically there are several examples of a given structure in an image to annotate to illustrate the various perspectives a 2D section captures of a 3D tissue structure. The same color and same anatomical name are assigned in this case. These examples are collectively merged in QuPath to collapse them into a single group annotation with a Class (Term) assigned. Next, several other structures are annotated using with the same approach with each group of like structures assigned a shared color and anatomy term after being merged. The resulting annotated image will have several sets of structures labeled and named for use.

Annotated BF image
Example of an annoated histological image using the GUDMAP-defined color palette.

In QuPath, colors for Classes are randomly assigned from a broad palette in the program. I sought to identify the most useful palette of colors for use in labeling histological images and applied the following 4 criteria:

  • Maximally distant colors
  • Accessibility measures: account for colorblind users
  • Avoid collisions with digitally restrained images
  • Colors can be seem on BF H&E and IF images

I started with a set of 64 max distance colors algorithmically derived for use in data figures (blog post - (Tatarize, 2012)). This panel was examined on a colorblind simulator and problematic colors removed from that set manually (tool - “Coblis — Color blindness simulator – Colblindor”). Since the goal here is to annotate histological images, emerging techniques that “digitally re-stain” images were considered to removed incompatible colors (an example is presented in Kather et al. 2015. Finally, colors close to the common H&E colors were tested for viewability and those that are difficult to see where removed.

This resulted in a list of 26 colors to use on H&E images that should also be compatible with most IF images. Annotation Color Palette
Image that includes the colors and the hex IDs for each.

The hex codes in text format:
"#D5FF00", "#00FF00", "#FF937E", "#91D0CB", "#0000FF", "#00AE7E", "#FF00F6", "#5FAD4E", "#01D0FF", "#BB8800", "#BDC6FF", "#008F9C", "#A5FFD2", "#FFA6FE", "#FFDB66", "#00FFC6", "#00B917", "#BDD393", "#004754", "#010067", "#0E4CA1", "#005F39", "#6B6882", "#683D3B", "#43002C", "#788231"

References

Coblis — Color blindness simulator – Colblindor. (n.d.). Retrieved from https://www.color-blindness.com/coblis-color-blindness-simulator/

Kather JN, Weis C-A, Marx A, Schuster AK, Schad LR, Zöllner FG (2015) New Colors for Histology: Optimized Bivariate Color Maps Increase Perceptual Contrast in Histological Images. PLoS One 10: PMCID: PMC4696851 https://doi.org/10.1371/journal.pone.0145572

Tatarize. (2012, September 6). Color distribution methodology. Retrieved from https://godsnotwheregodsnot.blogspot.com/2012/09/color-distribution-methodology.html