Skip to content
ProtocolProtocol
← Your pathLevel 1 · Beginner

Imaging - Bio-Formats, metadata, and measurement you can defend

Lesson 5 of 6 · Your instrument, end to end

1 · Learn the move · Verified-data override

PARALLEL TRACK - complete only if pl1 pointed you here. Imaging's one indispensable tool is Bio-Formats: a single converter that reads every vendor microscopy format - .czi, .nd2, .lif, .oib - and writes OME-TIFF; most bench scientists have never heard of it. The pipeline is conversion plus one audit, and the metadata IS the experiment: pixel size, channel identity, dimension order live in the OME-XML, and if conversion drops the pixel size, every measurement becomes unitless and the numbers keep looking fine. So extract the metadata before and after conversion, diff them, and reason only over that verified record, never over typical Zeiss values. Two hard rules finish the track: confirm dimension order before slicing any stack, and no number ever comes off a screenshot - quantification happens in Fiji, on the raw file, with the background method declared.

Author my imaging conversion-and-audit pipeline: (1) bfconvert command from [.czi/.nd2/.lif] to OME-TIFF; (2) a metadata extraction script (showinf or tifffile) run on BOTH the original and the converted file, pulling pixel size X/Y/Z with units, dimension order, channel names, objective; (3) a diff that fails loudly if pixel size or channel identity changed or vanished; (4) a measurement plan for [what I measure] that runs in Fiji on raw data with the background method declared - and a standing refusal: no quantification from any screenshot, export, or downsampled image. Reason only from the extracted metadata I paste, marking gaps [NEEDS: ...], never from typical values for my instrument.

2 · Your turn. You write the prompt

Ninety Zeiss .czi z-stacks from a colocalization study need to become OME-TIFFs your analysis scripts can read, and the one thing that cannot happen is the pixel size silently not surviving - your whole measurement is distances. Build the convert-audit-measure pipeline, with the metadata diff as the gate.

Remember: the AI sees only your prompt, not this page. If the situation isn't in your prompt, it doesn't exist.

Optional. These shape the output when you run your prompt below, not your score.