A 4 µm section is rarely 4 µm — thickness varies widely and shows in the H&E

Design

Experimental measurement study with AI model development

Population

154 measurement sites; 357 image–measurement pairs, 56-image test set

Primary outcome

Section thickness vs microtome setting; AI inference accuracy

Effect

>2/3 outside ±10% of setting; model MAE 0.28 µm, R² 0.86

This Laboratory Investigation study, published 17 September, measured histological section thickness directly with confocal surface profiling at 154 sites and matched it to H&E images, then trained AI models to infer thickness from the image alone.

More than two-thirds of paraffin-embedded measurements fell outside ±10% of the microtome setting. After deparaffinisation, sections shrank to about one-third of their embedded thickness, and shrinkage varied by tissue component — collagen, mucin, red-cell-rich areas and nuclei — producing marked unevenness within a single section. The best model (ResNet50) estimated thickness from H&E with a mean absolute error of 0.28 µm (R² 0.86) in an independent test set.

Section thickness changes nuclear density, staining intensity and apparent chromatin, which matter for both human reading and digital algorithms. It is an under-recognised pre-analytical variable, and one that could be checked automatically.

• Do not assume sections match the microtome setting; audit thickness periodically.
• Consider section thickness when a slide looks unusually hyperchromatic or pale.
• Standardise microtome maintenance and technique across technical staff.
• Expect thickness variation to affect digital pathology algorithms as well as eyes.

Why it matters

Explains some slide-to-slide variation in appearance that is usually blamed on staining.

The statistics, in plain English

An R² of 0.86 means the model's estimates tracked measured thickness closely; an error of 0.28 µm is small relative to a typical 3–4 µm section.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.