ViabiLens Uses Brightfield Imaging for Label-Free Cell Viability Analysis
Published in Artificial Intelligence, Biology, Tools.
October 8, 2026
A preprint submitted to arXiv on October 7 introduces ViabiLens, an AI-assisted approach to counting cells and assessing their viability from unstained brightfield microscope images.
Amir Reza Vazifeh and colleagues combine a cell detection model with a convolutional neural network that predicts whether individual cells are alive or dead. An interactive UMAP viewer helps researchers inspect cells across a sample.
The authors evaluated Chinese hamster ovary cells under varied viability conditions. They reported a mean absolute error of 2.68% on unstained samples compared with fluorescence-based reference measurements, and released a benchmark dataset.
Label-free assessment could reduce reliance on stains that disturb cellular processes or complicate continuous monitoring in cell culture. The reported results concern the evaluated CHO samples; they do not establish equivalent performance across other cell types or manufacturing systems.
This is a research preprint, not a peer-reviewed clinical validation or evidence of a cleared diagnostic product.
Source: Vazifeh and colleagues, arXiv:2610.10473, submitted October 7, 2026.