How to Quantify IHC Staining

Decide what to measure

Pick the measure before you look at the results, and use the same one for every sample. For a count-based index, the Ki-67 index calculator shows how far a count of a given size can be off.

Software for IHC quantification

A wider comparison of analysis tools is on microscopy image analysis software compared.

How to quantify IHC in QuPath

1. Open the slide and set the image type to Brightfield (H-DAB), so QuPath separates haematoxylin from DAB.
2. Draw an annotation around the tissue to score. Avoid scoring the full image: folds, edges and background skew the result.
3. Run Analyze › Cell detection › Positive cell detection on the annotation. Check that the detected nuclei match the image, and adjust the detection settings if nuclei are merged or missed.
4. For a nuclear marker set Score compartment to Nucleus: DAB OD mean. A single threshold splits cells into positive and negative; up to three thresholds grade them weak, moderate and strong (1+, 2+, 3+).
5. Read Positive % for the annotation in the Annotations tab. With three thresholds QuPath also calculates the H-score, from 0 to 300.
6. Open Measure › Show annotation measurements, export the table to CSV for your statistics, and note the thresholds in your methods.

The same intensity thresholds can be changed afterwards with Classify › Object classification › Set cell intensity classifications, without detecting the cells again.

How to quantify DAB staining in Fiji (ImageJ)

1. Open the brightfield RGB image. A whole slide is usually too large to open at full resolution, so export a region or field from it first.
2. Run Image › Color › Colour Deconvolution and choose the H DAB vectors. You get three images: haematoxylin, DAB and a residual.
3. For stained area: on the DAB image set a threshold with Image › Adjust › Threshold, tick Area fraction and Limit to threshold under Analyze › Set Measurements, then Analyze › Measure.
4. For percent positive cells: count nuclei on the haematoxylin and DAB images (threshold, Process › Binary › Watershed, Analyze › Analyze Particles), then divide positive nuclei by all nuclei.
5. Use the same threshold values for every image, and record the steps as a macro so the batch is repeatable.

The plugin's author warns against reading DAB darkness as an amount of antigen: DAB does not follow the Beer-Lambert law, and immunostaining is not stoichiometric. Compare intensity only within one staining run, and prefer counting positive cells or measuring area to comparing raw darkness.

How to score IHC in SlideScope (no code)

1. Open the slide in SlideScope on Windows or Mac (SVS, NDPI, MRXS, CZI, TIFF and the other formats it reads) and draw or zoom to the region to score.
2. Open Quantify, set Objects to nuclei and Stains to Hematoxylin + DAB (IHC). If the colours of your stain differ from the defaults, estimate the stain vectors from the slide.
3. Under Marker classification, turn on Classify marker-positive objects with DAB as the positive marker. The three grade thresholds (1+, 2+, 3+, in optical density) start at 0.2, 0.4 and 0.6 and can be edited.
4. Run it. Each nucleus is outlined on the image, so you can check the result, correct the class of a nucleus or exclude anything that was misread.
5. Read percent positive, H-score, Allred score and positive cells per mm² for the region, then export the object table (CSV), the outlines (GeoJSON, which QuPath opens) and a methods summary.

Or type the question. With Ask Your Slide you can ask “what percent of nuclei are Ki67-positive in this region?” or “give me the H-score and Allred score for the area I outlined”, and SlideScope runs these same tools on your computer and shows the evidence for every number. The scores use the same definitions as QuPath, so the two agree on the same cells. Scoring grades nuclei, which suits nuclear markers; results are for research use and need your visual check. More on the counting page.

What decides whether the number is right

• Where the marker sits: nucleus-based scoring suits nuclear markers such as Ki-67, ER, PR and p53. Membrane markers (HER2, PD-L1) and cytoplasmic markers need the stain measured in that compartment, and clinical scoring of them follows the assay's own manual.
• Thresholds set on controls: fix the positive threshold on control tissue from the same staining run, before scoring the experiment, and never adjust it per slide.
• The region: score the tissue the question is about, and leave out necrosis, folds, edge staining and stroma unless the protocol includes them. Hot spots and whole sections give different numbers; say which you used.
• Check the detection: look at the outlines. Merged nuclei lower the count and split nuclei raise it; crowded tissue needs a deep-learning detector or adjusted settings.
• Enough cells, the right n: score several hundred cells per sample, and run statistics on one value per animal or patient, not on cells or fields.
• Report it: software and version, stain vectors, thresholds, the region rule, cells counted and who checked the result.

Questions

Sources

• QuPath documentation: cell detection, Positive cell detection and Score compartment
• QuPath documentation: cell classification, intensity thresholds, Positive % and H-score
• ImageJ documentation: Colour Deconvolution (Ruifrok AC, Johnston DA. Quantification of histochemical staining by color deconvolution. Anal Quant Cytol Histol 2001;23:291-299)
• Landini G. Colour Deconvolution 2: the H DAB vectors, and why DAB darkness is not a measure of antigen
• Nielsen TO et al. Assessment of Ki67 in breast cancer: updated recommendations from the International Ki67 in Breast Cancer Working Group. J Natl Cancer Inst 2021
• Wen Z et al. Deep learning-based H-score quantification of immunohistochemistry-stained images. Mod Pathol 2023

This guide is for research and teaching. It is not a diagnostic procedure: scoring for patient care follows the validated assay, the current guideline for that marker and a pathologist's judgement.