The problem — Microscopy metadata is most accurate at the moment of acquisition and most commonly recorded days later, from memory, in a free-text methods section. By then, the objective numerical aperture, the exact laser power, the filter set, and the pixel-size calibration have either been forgotten or reconstructed imprecisely. Retrospective metadata is not provenance — it is an approximation of provenance, and the difference is measurable in downstream reproducibility.
What it is / how it works — Micro-Meta App is an open-source, browser-deployable tool developed by Rigano and colleagues (Nature Methods, 2021) that operationalises the 4DN-BINA-OME tiered system of Microscopy Metadata specifications. It presents instrument operators with an interactive interface that walks through the exact fields required by the community spec: hardware components (objective, laser, detector, filter sets), acquisition settings (pixel size, z-step, exposure time, channel assignments), and sample context. Each field maps to the OME data model, so the output JSON can be embedded directly in an OME-NGFF store or handed to a deposition tool like the BioImage Archive. The tool was validated across 16 core facilities to produce more uniform metadata records than lab-notebook practices. It addresses both tiers of the metadata framework that Metadata & Provenance: The Run Manifest draws on: provenance metadata (MPM — how the image was made) and quality metadata (MQM — how good it is).
Where it breaks — Micro-Meta App captures what an operator tells it; it cannot infer fields from a vendor file it has not been given. If the microscope session ends before the app is run, the value is lost to memory. The practical discipline is to treat Micro-Meta App as a concurrent step at acquisition — run alongside the vendor software — not as a post-hoc documentation tool. A half-filled record is better than none, but the acquisition-level fields (pixel size, objective NA, channel identities) are the ones that feed downstream validation; incomplete there means uncomputable downstream.