IHC scoring
Ki-67 index calculator
Enter the Ki-67-positive nuclei and the nuclei you counted to get the labelling index as a percentage, with a 95 % interval that shows how far the count alone can be off. A second calculator tells you how many nuclei to count for the precision you need. The formula, counting methods and their sources are below.
How the Ki-67 index is calculated
Ki-67 is a nuclear protein present in cells that are in the cell cycle and absent from resting cells, so the share of tumour nuclei that stain for it is a measure of proliferation. The Ki-67 index (also called the labelling index or proliferation index) is that share as a percentage.
Ki-67 index (%) = positive tumour nuclei ÷ (positive + negative tumour nuclei) × 100
Worked example: 180 brown (positive) and 820 blue (negative) tumour nuclei are 1,000 nuclei in all, so the index is 180 ÷ 1,000 × 100 = 18 %. A count of 1,000 nuclei can only pin that down to about 15.7 % to 20.5 %; the same 18 % from 200 nuclei could be anywhere from about 13 % to 24 %.
Any nuclear staining above background usually counts as positive, whatever its strength; only tumour nuclei are counted, not lymphocytes, stroma or endothelium.
Where to count: whole section or hot spot
Ki-67 staining is rarely even, so where you count changes the answer more than the arithmetic does. Use the method your tumour type's guideline names, and say which one you used.
- Global (breast cancer): the International Ki67 in Breast Cancer Working Group recommends a standardised visual method that scores fields representing the range of staining across the section and records a weighted global score, rather than the hot spot alone, which was more variable between observers.
- Hot spot (neuroendocrine tumours): grading reads the index in the area of highest labelling, with at least 500 tumour cells counted there, and many laboratories count from a printed or captured image rather than by eye.
- Eyeballing is the weakest method: estimates without counting vary most between observers. Counting on a captured image, or with image analysis you have checked, is more repeatable.
For the same working group, values of 5 % or less and 30 % or more were the ranges where Ki-67 was judged reliable enough to inform prognosis in early breast cancer; values in between agreed less well between laboratories. Clinical thresholds belong to the guideline for the tumour type, not to this page.
How many nuclei to count
A count is a sample, and its uncertainty shrinks with the square root of the number counted. The interval the calculator shows is the Wilson interval for a proportion. For planning, the usual normal approximation is enough: for an expected index p and a wanted margin of ± m (both as fractions), the nuclei needed for a 95 % interval are about:
n = 1.96² × p × (1 − p) ÷ m²
| Expected index | ± 5 points | ± 3 points | ± 2 points |
|---|---|---|---|
| 5 % | 73 | 203 | 457 |
| 10 % | 139 | 385 | 865 |
| 20 % | 246 | 683 | 1,537 |
| 30 % | 323 | 897 | 2,017 |
| 50 % | 385 | 1,068 | 2,401 |
This is the counting error only. It says nothing about which fields you chose or where you drew the line between positive and negative, which is why a fixed method matters more than a very large count.
Getting the same score from software
Image analysis software reaches these scores by the same arithmetic, with the eye replaced by a measurement. It finds every nucleus in the region, separates the brown DAB stain from the blue haematoxylin by colour deconvolution, measures the mean DAB optical density of each nucleus, and grades it against thresholds you set: one threshold for positive or negative, three for 1+, 2+ and 3+.
- QuPath (free, open source): Positive cell detection with one threshold gives percent positive; with three thresholds it also reports the H-score.
- Fiji / ImageJ (free): Colour Deconvolution with the H DAB vectors, then threshold and count; the scoring arithmetic is yours to do.
- SlideScope: counts nuclei in the region you choose, grades each one, and reports percent positive, H-score and Allred score together, using the same definitions as QuPath. You can also ask in plain language with Ask Your Slide; every number comes from the local tool, with outlines to check. Research use only.
Whichever you use, the thresholds are choices: write them in your methods, check the outlines against the image, and keep them fixed across the slides you compare. Step by step: how to quantify IHC staining.
Questions
How do you calculate the Ki-67 index?
Divide the number of Ki-67-positive tumour nuclei by all the tumour nuclei counted (positive plus negative) and multiply by 100. With 180 positive out of 1,000 nuclei the index is 18 %.
How many cells should be counted for Ki-67?
It depends on the guideline for the tumour type and the precision you need. Neuroendocrine tumour grading counts at least 500 tumour cells in the hot spot. Statistically, about 700 nuclei give ± 3 percentage points at an index near 20 %, and about 250 give ± 5.
Should Ki-67 be counted in hot spots or across the whole section?
Follow the guideline for the tumour type. The International Ki67 in Breast Cancer Working Group recommends a weighted global score across the section; neuroendocrine tumour grading uses the hot spot. Always state which you used.
What is a high Ki-67 index?
There is no single cut-off: it depends on the tumour type and guideline. In early breast cancer the International Ki67 Working Group judged 5 % or less and 30 % or more reliable enough to inform prognosis. For a patient result, ask the reporting pathologist or treating doctor.
Can software count Ki-67?
Yes. Image analysis detects every nucleus in a region and classifies it as positive or negative by its DAB staining; QuPath and SlideScope both report percent positive this way. The working group describes automated scoring as still investigational for clinical use, so check the outlines and thresholds and treat the result as research output.
Is this calculator free?
Yes. It runs in your browser and sends nothing anywhere. SlideScope, the desktop microscopy and pathology app made by the same team, is a paid app with a free sample library (every tool) and a subscription for your own images.
Sources
- 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;113(7):808-819
- Trikalinos NA et al. Accuracy of grading in pancreatic neuroendocrine neoplasms and effect on survival estimates. Ann Surg Oncol 2020 (hot-spot counting of at least 500 cells)
- QuPath documentation: Ki67 labelling index by positive cell detection
- NIST/SEMATECH e-Handbook of Statistical Methods, 7.2.4.1: confidence intervals for a proportion (the Wilson method)
These calculators do arithmetic for research and teaching. They are not a diagnostic tool: scoring for patient care follows the validated assay, the current guideline for that marker and a pathologist's judgement.
Score the slide instead of counting by hand
SlideScope counts the nuclei in a region of your slide, grades each one by its DAB staining, and reports percent positive, H-score and Allred score with the outlines to check, on your own computer. Or type the question: Ask Your Slide runs the same tools and shows the evidence. Research use only. Try every tool free on the built-in samples; opening your own images needs a subscription. Already decided? Subscribe now.