Pipeline stage

Features / Embeddings

Transform segmented objects into numerical representations — hand-engineered morphology descriptors or self-supervised learned embeddings — that encode the phenotypic state a biologist reasons over.

The problem — A segmentation mask is not a measurement. The features stage is where the biology becomes computable: each object is reduced to a vector, and that vector determines which phenotypes are detectable at all. A feature space that cannot represent a particular morphological change means that perturbation is invisible to the entire downstream analysis — not weak, invisible.

What it is / how it works — Two families. The classical profiling family (CellProfiler lineage) computes hundreds of named morphology, texture, and intensity descriptors per object — the backbone of the Cell Painting assay, which yields ~1,500 features per cell across five channels (Bray et al. 2016; decade review: Seal et al. 2025). Each feature has a biological reading; the profile is interpretable and auditable. The learned embedding family trains an encoder to emit a compact representation, either supervised (foundation models for pathology: UNI, H-optimus-0, CONCH) or self-supervised (DeepProfiler, scDINO, cytoself). Self-supervised models — particularly those in the VQ-VAE or ViT lineage — can out-cluster classical features on organelle and complex membership, as cytoself demonstrated on protein localization. The decade review reports deep-learning features beating CellProfiler by up to +29 mAP on JUMP-scale benchmarks.

Where it breaks — Embeddings buy sensitivity and lose interpretability. A learned axis that separates two compounds tells you that they differ, not how — and it encodes acquisition nuisance (batch, scanner, plate position) just as readily as biology. Self-supervised models are notorious for batch-encoding; cytoself's authors flag residual batch effects. The features stage never validates itself: whatever vector is produced must clear the correction and reproducibility gates before any biological claim is credible. See From Masks to Morphology: Features & Self-Supervised Embeddings for the full trade-off.

A higher-dimensional embedding is not a better one until you have shown that the added variance is biology, not optics or batch.

References

Appears in these notes

  • From Masks to Morphology: Features & Self-Supervised EmbeddingsOnce cells are segmented, you turn masks into numbers — either hand-engineered morphology features or learned embeddings. Each choice trades interpretability against the signal it can capture.
  • Cell Painting Pipelines — From Images to Phenotypic ProfilesA Cell Painting profile is the output of a long pipeline — illumination correction, segmentation, feature extraction, batch correction, then a reproducibility readout — and the number that matters (percent-replicating / mAP) lives only at the end. Every stage upstream can silently degrade it.
  • Choosing a Bioimage Analysis Pipeline — Off-the-Shelf vs CustomMost microscopy questions are answered by composing existing, validated tools — and a custom pipeline is justified only when off-the-shelf options fail on a specific, identifiable axis. This note gives the decision criteria and the signals that you have genuinely crossed into custom territory.
  • Evaluating Segmentation Models in Real Microscopy WorkflowsA segmenter's benchmark Dice rarely predicts its behavior in a real workflow. Instance counting, merge/split errors, and cross-instrument generalization are the quantities that decide whether the downstream readout is correct — and standard semantic overlap scores are blind to all three.
  • Foundation Models Are Not PipelinesA foundation model is a powerful feature extractor, not a deployable measurement pipeline. It still needs ingestion, QC, preprocessing, confounder correction, validation, and provenance around it — and a frozen encoder makes the surrounding system more important, not less.
  • Ground Truth Is a Design ProblemGround truth in microscopy is constructed, not given — every label is the output of an annotation protocol, an annotator, and a fusion rule, each with its own bias and variance. Treating it as a fixed answer key is how inter-rater noise and label bias get baked into every score computed against it.
  • 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.
  • Model Zoos for Bioimage Analysis — From a Zoo Model to a Validated WorkflowThe BioImage Model Zoo and community packaging tools make pre-trained models shareable, runnable, and reproducible across tools — solving distribution and provenance. They do not solve fitness: a downloaded model is a component, and the path from a zoo entry to a trustworthy readout still runs through validation.
  • The Bioimage Analysis Ecosystem — Where Each Tool FitsBioimage analysis is not one tool but a landscape organized along two axes — interactive vs scriptable, and generalist vs specialist. Fiji, napari, CellProfiler, QuPath, ilastik, and the deep-learning segmenters each occupy a different cell of that grid, and choosing well means knowing which cell your problem lives in.
  • Validation Without a Ground TruthValidation asks whether a microscopy pipeline's outputs are biologically true and fit for purpose — but biology rarely supplies a clean answer key. The discipline is choosing metrics that reflect the question, manufacturing ground truth honestly, and conditioning performance on the experiment.

← Back to the constellation