CellCount specifications.

What CellCount reads, what it returns, and what it runs in. Every figure here is the one printed on the validation page, and every setting is the one the product offers.

Images in

File formats

JPEG, PNG, TIFF

HEIC and HEIF photos are converted to JPEG in the browser on Safari and iOS, the only browsers that can decode them; elsewhere, export the photo as JPEG first.

Resolution

The camera’s full resolution

A frame needs 1,650 px on its short side to count a 1 mm square, and proportionally fewer to count a smaller one: 330 px for a 0.2 mm square. Images are uploaded as captured: nothing is resized on the way in, because resizing moves the count and compression does not.

Chamber

Improved Neubauer hemocytometer, bright-line ruling

The whole of what one image counts (one 1 mm square, unless the image counts smaller ones), with all four corners, in the frame. The rulings in the image are what the scale is measured from.

The grid

Found automatically in each image; if the rulings cannot be found, the crossings are placed by hand

The rulings give the scale, the square that is counted and so the volume. The detector reads the 1 mm squares and the 0.2 mm medium squares of the centre. A frame with no ruling, such as one an automated counter exported, takes its scale from the file.

Stain

Trypan blue

Clear cells are counted live, blue cells dead.

Focus and exposure

Rulings crisp; detail kept in both the bright and the dark areas

Operating envelope

Scale at the specimen

Rated from 1,500 px/mm; 1,200 to 1,500 px/mm is a caution band

Measured in every image from the chamber rulings, not taken from the objective or the camera’s settings. Each image is shown with its scale and where it sits: in range, caution or out of range.

Smallest cell

10 µm at the rated scale; 12.5 µm in the caution band

Squares counted

The 1 mm squares and the 0.2 mm centre squares are found automatically; the 0.25 mm corner squares are marked by hand

If the rulings cannot be found, the crossings are placed by hand. A smaller square still fits the frame at a higher magnification, which suits dense samples.

Largest scale

Whatever still fits the counted square on the photo’s short side

For one image: short side ÷ square side, in px/mm. When choosing a camera and reducer, leave a 5% margin on each side: short side ÷ (1.1 × square side).

Objectives

Rated at 4× and 10×

Higher objectives frame the smaller squares; they are not rated until they have been validated.

Highest validated scale

2,690 px/mm

The highest scale among the reference captures the envelope was validated on. A setup above it can frame its square, but its accuracy has not been characterised yet.

Two frames side by side. Left, choosing hardware: the counted square plus a margin on each side has to fit the photo's short side. Right, judging one image: once the square has been found on the image, the square itself only has to fit.
The framing bound. A camera and reducer are chosen with a margin around the counted square; one image only needs the square inside the frame once the square has been found on it.
Log-log plot of scale at the specimen against the camera's short side. A 12 megapixel reference camera is drawn as four rails, one per objective from 4x to 40x, each marked at every C-mount adapter from 0.35x to 1.0x. Diagonal lines are the framing limits for a 1 mm, 0.25 mm and 0.2 mm square; coloured regions mark in range, caution, under-resolved, and too magnified for a 1 mm square or for every square.
A microscope camera: where each objective and C-mount adapter puts the scale on a 12 MP reference camera, against the framing limit of each square size.
Log-log plot of scale at the specimen against the photo's short side. Vertical lanes mark the photo sizes common phones take, from 12 megapixel defaults to 200 megapixel modes. Diagonal bands show a phone at the eyepiece at 10x, 20x and 40x; only the 10x point was measured, the rest is projected. Coloured regions mark in range, caution, under-resolved, and too magnified for a 1 mm square or for every square.
A phone at the eyepiece: where 10×, 20× and 40× land against the photo sizes common phones take. Only the 10× point was measured; the rest is projected from it.

Ways in

A camera on the microscope port

A USB microscope camera, counting live in the browser

The recommended setup. Ask us for a plug-and-play camera matched to your scope.

A phone at the eyepiece

A photo from a modern phone held to the eyepiece; no adapter

Files you already have

JPEG, PNG, TIFF files, one at a time or a whole folder

Including the raw frames an automated counter exports. Raw frames, not the instrument’s overlay images.

The sample

Images per sample

Up to 150 images of one biological sample

Results are given per image and pooled for the sample.

The protocol

Dilution

Any factor; presets up to ×20; one part sample to one part trypan blue is the default

Counting pattern

Which squares are counted: any of the nine 1 mm squares, all nine in one image, or the 0.2 mm medium squares of the centre, one or a 2 x 2 block in each image

Sensitivity

Low, Balanced, High

High for small cells such as yeast; Balanced for most samples, including mammalian cells; Low when only confident detections should count.

Cell-size gate

A minimum and a maximum equivalent diameter in micrometres, or off

Sharing

Anyone on the team can create a protocol; every count in the lab follows the one it was made under

The same protocol runs on a chamber image and on a counter’s exported frames.

Versioning

Protocols are versioned; a count records the version it followed

What comes back

Per image

Live and dead counts, with every detected cell marked on the image

Concentration

Cells per mL, from the grid measured in the image and the dilution

Viability

The live share of counted cells, in percent

Size

A cell-size histogram pooled over the sample, with the gate drawn on it

Corrections

Add a missed cell, delete a wrong one, flip live to dead, re-run at another sensitivity

Nothing is final until the person who knows the sample accepts the count.

The record

Acceptance

By a named, signed-in user (their email), at a stated time

Contents

The counts, the grid, the dilution, the protocol with its sensitivity and gate, the model version, the acceptor and the time

Integrity

A SHA-256 digest computed over the record’s exact bytes, stored with it and served with it

Anyone holding the record later can recompute the digest and confirm nothing changed since acceptance.

Documents out

Results document

Sample result, size distribution, per-image table, method, software, statistical analysis, and an appendix of the counted frames with their overlays

Names who accepted the sample and when. Opens in the browser and prints to PDF.

Export bundle

A zip with the original images, the annotated images, the canonical frames and the per-image results as CSV, plus a manifest and a README

For the lab’s own archive; the count can be reviewed later without the app.

Agreement with a leading automated counter

Model

Our released Huginn 2.0 trypan blue exclusion model

Every figure in this group is for that model.

Lin’s concordance

0.91 on cell concentration, over 16 images CellCount had never seen from that instrument

Bland-Altman

Mean difference 0.002 million cells/mL; 95% limits of agreement ±0.29 million cells/mL

Live-or-dead call

The same call on 95% of the 6,765 cells both readers found

The method, the data and every number are kept with the study and are available on request.

Runs in

Browser

Current Chrome, Edge and Firefox; Safari 15 or newer; iOS 15 or newer

Needs WebGL2. Nothing to install.

Devices

Laptop or desktop at the bench, tablet, or phone

Terms

How data is handled is set out in the Privacy Notice and the Terms of Service, and institutions can read the Enterprise Terms. The legal entity on every agreement is Lowkey Vision AS.

Try the live demoCell counting terms, definedAsk about your setup