Image Analysis Quality Control

Part 3: Staining QC

Image Quality Control (QC) is an essential first step in any image analysis study. At OracleBio, we QC check every image before analysis and only proceed with images that meet certain criteria, taking into consideration 1) Image Scan 2) Tissue Section and 3) Staining. In the final of our 3-part Image Analysis Quality Control Blog series, we discuss what to look for when reviewing the staining of a sample to determine suitability for analysis.

Staining:

It is important to review the staining of the sample and note if there are any irregularities present that could impact analysis, such as background staining, edge effect or high staining variability across the sample.

‘Edge effect’ is the term used for the staining pattern observed when the outside of a tissue preferentially takes up staining more than the inside, resulting in a ring of non-specific differential staining around the edge of the tissue. These areas should be excluded prior to analysis as they can result in false positive or inaccurate data readouts.

If there is background staining present, antibody isotype controls can be useful for defining the difference between specific and non-specific staining. For staining that uses secondary antibodies, background staining can be defined by control tissue in the absence of  the primary antibody.

Staining variability across the sample can be caused by a number of errors such as uneven tissue fixation, application of reagents, issues with the automated stainer or by failing to keep the slides flat until they have dried. Affected areas can often be excluded during annotations however high staining variability can result in an image failing QC.

Right: Examples of edge effect in an H&E stained section

Nuclear Features:

In image analysis, the identification of a cell is often driven by the clear identification of a nucleus.   If the nuclei staining is indistinct, as can happen in some densely packed cellular structures, then formation of cell objects and cell segmentation during analysis may be compromised. Capturing the extent of this and its impact on the analysis is important to the interpretation of data sets that come from the analysis.

Left: Example tissue areas with indistinct nuclei

Target Staining:

Staining that is too weak or too strong can produce false negative or false positive results. For example, a dark blue overstained haematoxylin nucleus may be difficult to differentiate from a dark brown DAB labelled positive nucleus, or even mask weakly DAB labelled positive nuclei. In addition, morphological characteristics can be difficult to discriminate by inappropriate stain levels. If the target is strongly over or under stained then this can mean that an image is not suitable for analysis.

Right: Overstained Tissue Section

Accurate annotations – Defining the regions of interest:

The differentiation between tumour and normal tissue or tumour and stroma is an important first step in the analysis of tumour tissue samples. Pan-Cytokeratin (PanCK) is regularly used as a tumour identification marker. However, it is not specific to tumour alone and areas of epithelial structures/ducts can also take up the PanCK stain. Tumour generally has larger, randomly assorted cells (a progression from pre-cancerous dysplasia) whereas epithelial structures tend to have smaller, more uniform cellular morphology. It is important to review PanCK staining during image QC to ensure accurate definition of the tumour ROI.

Left: PanCK identifying non-tumour areas

There are many factors to consider when deciding if an image is suitable for image analysis and these are just some to include when conducting Staining QC. To find out what else we include in our QC check, click here to read Image Analysis Quality Control Part 1: Image Scan QC and Image Analysis Quality Control Part 2: Tissue Section QC.

Encountering some of the issues above does not necessarily mean that an image cannot be analysed as some problems can be overcome, for example by negative annotation or additional algorithm development. Unfortunately, at OracleBio we do have to fail a significant number of images every year because the image analysis data generated will not be robust. If you are unsure if your images are suitable for quantitative analysis or need advice on how to avoid encountering some of these issues, please contact us.

About the Author:

Nicole Stillie

Nicole joined OracleBio in 2016 after graduating with a B.Sc. in Biomedical Science. She is now a Senior Image Analysis Scientist and QCs hundreds of images every month. Connect with Nicole on LinkedIn. 

To keep up to date with all OracleBio news and blogs, please follow us on LinkedIn and Twitter:

Find us on