Clarke & Treanor 2017 — Colour in Digital Pathology: A Review

A review of why colour varies across the whole digital pathology chain, from stain to scanner to monitor, and what calibration can and cannot fix.

A review arguing that colour variation in pathology was tolerated for a century under the optical microscope, and that digitisation both adds a new layer of variation and, for the first time, makes the whole chain measurable.

The argument

Colour is not decoration in pathology — it is the signal, because histochemical and immunohistochemical stains are how tissue features are made visible at all. The review’s observation is that the profession evolved alongside wide slide-to-slide colour variation “with little investigation into the implications”, because a pathologist at an eyepiece adapts to it without noticing.

Digitisation changes that in two directions at once, and this is the part worth carrying forward:

1. The digitisation process itself introduces further colour variation — scanner optics, illumination, sensor and the display all sit between the stain and the eye, and each has its own colour behaviour.
2. Image analysis algorithms use colour or intensity measures to detect and measure tissue features, so variation that a human would silently compensate for becomes a direct source of algorithmic error.

The review notes FDA guidance stating the need to control colour reproduction throughout the digitisation process for whole slide imaging in primary diagnostic use — i.e. this is a regulatory expectation, not only a research interest.

Why it matters here

This is the paper that frames the distinction the group’s own tooling keeps running into: a colour problem can be attacked physically at the scanner, or computationally after the fact, and they are not the same intervention. The applied work in this wiki has mostly been the second kind; see Colour Calibration for the split and Stain Normalisation for the software side.

Provenance note

Authors are Leeds — Leeds Teaching Hospitals NHS Trust and the Leeds Institute of Cancer and Pathology. Treanor declares advisory-board membership with Sectra and Leica/Aperio and a patent assigned to Roche-Ventana, all without personal remuneration. Worth recording because scanner-vendor proximity is directly relevant to a review about scanner colour behaviour; it is disclosed, not hidden.

Related: Colour Calibration — this review is the anchor for that page. Related: Scanner and Stain Variability — the same variation seen from the modelling side.