No cell left behind: automated physics-based tracking of {\em every} cell in a dense and growing colony

TL;DR

This paper introduces a physics-based simulation software that automatically tracks every cell in dense, noisy colonies, outperforming existing methods by mimicking human scene understanding.

Contribution

The novel approach uses physical rules to simulate cell activity, enabling comprehensive and noise-robust tracking of all cells in dense colonies.

Findings

Successfully tracked over 200 cells in a growing colony

Achieved robust tracking without extensive image processing

Demonstrated accurate lineage and movement analysis

Abstract

A human watching a video of closely-packed cells can generally identify every individual cell, regardless of density and noise, but most currently-available cell-tracking software cannot. This is because the human brain automatically builds a physical model of the scene as it progresses, allowing it to readily distinguish cells from noise and not be unduly confused by overlapping cells. Here we introduce software that uses physical rules to create a simulation of the activity in a cell video, synchronizing itself with the video as the activity progresses. Because our simulation includes every individual cell, we are trivially able to track all cell movement, growth, and divisions. Our method is also particularly robust to noise without requiring any substantial image processing. We demonstrate the effectiveness of this method by tracking the motion and lineage tree of a densely-packed colony of cells that grows from 4 to more than 200 individuals.