The impact of stain variation on automated cell count


Background: In pathology, differences between laboratory staining processes and reagents can produce slides that vary in their resulting stained color for nominally the same image. This has traditionally been accepted as a ‘normal’ level of variation within pathology, but with digital pathology variability in color may impact upon onward image analysis. Our previous research quantified H&E stain variation in eight UK laboratories using a novel, objective method of measurement. This work explores the impact that the measured variation had on automated nuclear counts on regions of liver tissue.

Methods: An automated nuclear count was applied to liver regions to quantify the impact of inter-laboratory stain variation. A range of stain normalization techniques were applied to the image regions to investigate if normalization techniques could mitigate the impact of stain variation on automated nuclear count. A selection of target images covering a range of stain intensities were used to investigate how this can affect the results.

Results: The results show that inter-laboratory stain variation affected automated nuclear counts. The stain normalization techniques used showed a wide range of results demonstrating a potential source of onward variability/uncertainty depending on the methodology used.

Conclusion: The stain normalization techniques used resulted in a wide range of cell counts.  Although this may not result in a clinical impact it does suggest standardising staining, rather than retrospective normalisation, may be a better approach for consistent computational analysis.



  1. Understand the levels of inter-laboratory H&E stain variation
  2. Understand that stain variation may impact upon image analysis, such as automated nuclear counts
  3. Understand that stain normalization techniques may not sufficiently 'normalize' images to mitigate the impact of variation on image analysis


Presented by:


Catriona Dunn, BSc

PhD Student

Leeds Teaching Hospitals NHS Trust


Catriona Dunn is a PhD student at the University of Leeds, and a digital pathology scientist working within the National Pathology Imaging Co-operative (NPIC) at the Leeds Teaching Hospitals NHS Trust, UK. Catriona studied Medical Sciences for her undergraduate degree and, shortly after completion, began working in digital pathology research. She started her PhD in 2019 and recently submitted her thesis describing the development of a novel H&E stain quantification technique, and investigating the effect of stain variation on image analysis.