Modality

Live-Cell Tracking

Fluorescence time-lapse imaging quantified as per-track dynamics — motility, signaling, division, and death — by linking segmented objects frame-to-frame across a stabilized movie.

The problem — A live-cell experiment that cannot reliably link the same cell across hundreds of frames produces a track table that is, in effect, a table of coincidences. Motility statistics, signaling kinetics, and division timing all depend on frame-to-frame identity being correct. Errors compound: a single missed link breaks the lineage; a spurious merge assigns one cell's fate to another. Most pipeline failures here are invisible — the track table is populated and plausible-looking, and the error only surfaces when a biological claim does not replicate.

What it is / how it works — Live-cell imaging captures fluorescence time-lapse movies: multiple positions, multiple channels, hundreds to thousands of frames. The pipeline moves in two passes. First, each frame is segmented independently — Cellpose or a comparable model produces per-frame masks and confidence scores. Second, a tracker links those masks across time, building continuous trajectories. TrackMate, btrack, and ultrack all approach the linking step differently (LAP assignment, Bayesian graphical model, network-flow optimisation), but all of them expose the same failure surface: missed detections, cell–cell occlusion, and rapid mitosis.

Before tracking can be trusted, two pre-conditions must hold. Frame registration (pystackreg) corrects for stage drift so apparent motion is real motion, not instrument jitter. Illumination correction (BaSiCPy) removes flat-field variation that would otherwise be read as intensity change over time — a direct confound for any signaling readout.

QC in this modality is temporal: focus and photon budget must hold across the full run, not just in the first frame. Phototoxicity and photobleaching are silent degraders — the cell is responding to the measurement, and the signal is decaying. A movie-QC report flags both.

Where it breaks — Track integrity, not segmentation accuracy, is the quantity that matters for biology. Per-frame segmentation metrics do not predict linking performance — a model that holds Dice steady while producing fragmented or merged objects at confluent regions will generate systematically wrong lineages. Division events are particularly vulnerable: mitotic rounding changes cell morphology sharply, confusing the linker at precisely the moment the lineage branches. The discipline is to validate track-level outputs against labeled lineage benchmarks, not just per-frame overlap, and to report phototoxicity, bleaching decay, and positional drift as first-class QC outputs alongside the event timeline.

Per-frame segmentation accuracy does not imply track integrity. Evaluate linking quality explicitly — missed divisions and identity swaps do not leave signatures in per-frame metrics.

How we run itthe Live-Cell Tracking pipeline, stage by stage — the nine-stage imaging backbone reshaped for this modality, with the tools, sub-steps and deliverable at each stage.

References

Appears in these notes

  • Light-Sheet Microscopy — The Pipeline Problem Behind the Beautiful VolumeThe rendered light-sheet volume is the easy part. Behind it is a petabyte-scale systems problem — ingest and chunking, deconvolution and destriping, stitching and registration, then tracking — and every stage is where the science silently degrades, not the final render.
  • The Challenge of Rare Events and MutantsWhen the phenotype you care about appears in 10 cells out of 300,000, accuracy is meaningless, validation has almost no positives to learn from, and false discovery dominates. Rare-event detection is a class-imbalance and ground-truth-scarcity problem — best answered by adaptive acquisition that goes looking for the positives.
  • From Research Script to Production PipelineA notebook that worked once is not a pipeline, and a community model that scored well on its paper's data is not a validated component. Turning either into something deployable is the engineering work of pinning, wrapping, gating, and validating — most of which the original artifact deliberately skipped.
  • Measure Where It MattersAdaptive, uncertainty-driven acquisition treats the microscope as part of the model — spending photons and time only where the image is uncertain or the biology is happening, instead of scanning everything uniformly.

← Back to the constellation