CountAnywhere: frugal cell counting and viability from smartphone hemocytometer photos

CountAnywhere accurately counts cells and scores trypan-blue viability from smartphone hemocytometer photos, achieving 98% sensitivity and 99% specificity and 0.999 concordance on 128 quadrants.

• Why it matters: Cell concentration and viability measurements are often slow, operator-dependent, or require dedicated automated hardware and disposable slides, limiting throughput and consistency.
• What they did: The browser-based app uses the hemocytometer grid in each image for scale, focus checking, and stereological counting, then segments cells with a Cellpose-SAM model fine-tuned on five cell lines.
• The result: Fine-tuning increased segmentation F1 from 0.46 to 0.97, and app counts matched ground truth better than manual counts (mean absolute error 5.3% vs 10.7%) without density-linked drift.

AI summary of the abstract · may contain errors · not medical advice · read the paper.