Pipeline stage

Report / API

Package the biological readout with its QC evidence and provenance lineage into a structured, deliverable artifact — an annotated report, a data export, or a queryable API — so results are reproducible, traceable, and safe to act on.

The problem — A pipeline that computes a correct result and packages it without QC evidence has done half the job. The consumer — a biologist, a project manager, a regulatory reviewer — sees the answer but not the evidence that it should be believed. This creates the exact failure mode Varoquaux & Cheplygina document for medical imaging: methodologically weak results that propagate because the evidence of their weakness was never surfaced.

What it is / how it works — The report stage is the accountability layer. It packages three things together: the biological readout (profiles, hit list, slide prediction, dynamics summary), the QC evidence at each of the four nested quality levels (image-space, run/sample, readout, decision — per the model in Image Quality Is Not One Number), and the provenance lineage that traces the result back to its acquisition. In practice this means: a Cell Painting run ships profiles plus a per-plate reproducibility table plus the image-QC flags that were applied; a digital pathology slide ships a grade plus per-tile QC plus the normalization and embedding versions used; a spatial omics dataset ships an AnnData export plus registration quality metrics. The delivery form — PDF report, structured JSON, API endpoint, interactive notebook — is secondary to the completeness of the evidence chain. This is the concrete form of The QC-Aware Report.

Where it breaks — The most common failure is reporting only the endpoint — the profile, the score, the count matrix — with no attached evidence of the QC checks that cleared it. This makes a result that should be flagged look identical to a result that is solid. A secondary failure is provenance detachment: a report that does not trace back to its acquisition manifest cannot be reproduced or compared across runs — the failure described in the metadata perspective (Huisman et al.). The discipline: the report is not done until it includes the decision ("reportable / needs-review / reject") and the evidence chain that supports it.

A report that shows only the result is asking for blind trust. A Fovea Lab report shows the result, the QC evidence, and the explicit decision — so a bad result is un-trustable on sight.

References

Appears in these notes

  • The QC-Aware ReportA QC-aware report doesn't just present a result — it surfaces the evidence that the result is trustworthy, structured by the four nested levels of quality and ending in an explicit decision. Its job is to make a number un-trustable on sight when it shouldn't be trusted.
  • Image Quality Is Not One NumberFovea treats microscopy quality as four nested levels — image-space, run/sample, readout, and decision — because a metric at one level says nothing about the levels above it. Full-reference scalars like SSIM and PSNR measure only the first, yet teams report as if they measured the third.
  • Why mAP, Dice, and SSIM Are Not EnoughDice, IoU, mAP, and SSIM measure overlap or similarity against a reference — not whether the result supports the decision the experiment exists to make. The discipline is problem-aware metric selection: derive the metric from the question, not from what is easy to compute.
  • Verification — A Pipeline You Can Re-RunVerification is the engineering discipline of making a microscopy pipeline deterministic, version-pinned, and re-executable from a provenance record. Without it, every result is a one-off, and "we changed nothing" is unprovable.
  • Every Output Should Know Where It Came FromProvenance is a property of the artifact, not the pipeline — every mask, profile, and count matrix should carry the lineage that produced it. Where verification asks whether a run reproduces, provenance asks whether a single output, found alone, can still explain itself.
  • How to Make Imaging Data Ready for ReanalysisReanalysis-ready means a third party with no contact with the original lab can re-run, re-segment, or re-interpret the data correctly — which requires FAIR principles, complete metadata, and deposition in a public archive. The bar is reuse by a stranger, not retrieval by the author.
  • 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.
  • Uncertainty, Drift, and Failure Modes in Bioimage AnalysisModels fail silently when the data drifts away from what they were trained on — and an overconfident, miscalibrated model gives no warning. Trustworthy deployment needs calibrated uncertainty, explicit distribution-shift detection, and a way to flag failure when there is no ground truth to check against.
  • Verification and Validation Are Two Different QuestionsVerification asks whether the pipeline was built right — deterministic, reproducible, correct to spec. Validation asks whether it is the right pipeline — outputs that are biologically true and fit for purpose. Microscopy makes both hard, and conflating them is how silently wrong results get shipped.
  • Why Microscopy Needs Verified OutputsA measurement you cannot reproduce is not a measurement. Why verification is a first-class pipeline stage, not an afterthought.

← Back to the constellation