Choose an image
PNG, JPEG, WebP, BMP or GIF, up to 30 megapixels. TIFF or HEIC: convert to PNG first.
Loading OpenCV...
Result
0.00%
of the image is microtubule
Share of all pixels counted as microtubule after the nucleus is removed. Background is in the denominator, so a tighter crop around the cell gives a higher number.
- Image
- 0 × 0 px
- Otsu threshold
- 0
- Microtubule pixels
- 0 / 0
- Nucleus removed
- 0 %
What these numbers mean
- Image
- Width × height; all of these pixels are the denominator.
- Otsu threshold
- Green brightness (0–255) above which a pixel counts as microtubule, chosen automatically from the image's own histogram.
- Microtubule pixels
- Pixels above the threshold with the nucleus removed and a small clean-up, over all pixels.
- Nucleus removed
- Share of the picture masked out as nucleus (blue) before counting.
Is that high or low?
| Condition | Microtubule share |
|---|---|
| Untreated | 27.9 ± 4.6 % |
| Nocodazole 45 µM | 17.2 ± 2.6 % |
| Taxol | 30.7 ± 9.5 % |
Those numbers only apply to pictures taken the same way: brightness, exposure, magnification and how tightly the cell is cropped all move the number, so compare images within one experiment rather than against this table.
How the number is computed
- Take the green channel of the image.
- Find the nucleus: Otsu threshold on the blue channel, then close small gaps (5 px ellipse, twice).
- Find microtubules: Otsu threshold on the green channel.
- Remove the nucleus from the microtubule mask.
- Clean up: open then close with a 3 px ellipse.
- Percentage = microtubule pixels ÷ all pixels × 100.
These are the same OpenCV steps as the Python pipeline, run by OpenCV.js in a background thread of your browser. On the three sample cells the browser result equals the Python script's result to four decimal places, and on a 60-image test set every picture matches within 0.01 point (see the accuracy table). Colour profiles embedded in PNG/JPEG files are ignored, exactly as the Python script ignores them, so Safari and Chrome give the same number.