parsing photometry from NWB files

[bendichter]

This dataset: https://dandiarchive.org/dandiset/001632 correctly uses the FiberPhotometry extension to store data, however this is not parsed by the DANDI CLI. Ideally, the filenames and asset and dandiset metadata would reference the fiber photometry technique.

[bendichter]

At least we want to express this in the assets summary metadata

[bendichter]

We now have dozens of fiber photometry dandisets and no way to mark them. Asset-level approach and measurementTechnique come entirely from neurodata_typemap in dandi/metadata/util.py, and the archive just aggregates those into assetsSummary. I scanned the assetsSummary of all 895 dandisets, and the only approach values in the whole archive are the four that map emits (electrophysiological, behavioral, microscopy, optogenetic), and the only measurementTechnique values are the twelve it emits. A search for "photometry" turns up 25 dandisets, but none of them carry any photometry-related metadata, so there is no facet or query that can find them.

Two places in the CLI would need to learn about the photometry types: neurodata_typemap, which drives the asset metadata, and _populate_modalities in dandi/organize.py, which drives the _ecephys-style filename suffix. The latter derives modalities by introspecting pynwb submodules, so extension types never get one.

One note on the example dataset: 001632 actually uses the older ndx-photometry extension, not ndx-fiber-photometry. Its files contain FiberPhotometryResponseSeries, FibersTable, ExcitationSourcesTable, FluorophoresTable, and PhotodetectorsTable, while ndx-fiber-photometry uses FiberPhotometryResponseSeries, FiberPhotometryTable, OpticalFiber, ExcitationSource, Photodetector, and Indicator. We should cover both namespaces.

@yarikoptic since whatever terms the CLI emits effectively become the archive vocabulary, do you have a preference here? I would propose "fiber photometry approach" and "fiber photometry technique", following the pattern of "optogenetic approach". For the organize filename suffix I do not have a strong precedent to point to (I do not see an established BIDS suffix for fiber photometry), so _fib or _fiberphotometry both seem reasonable to me.

[yarikoptic]

I think this is the PR we should finish (with @rly HI or some AIs):

• Add ndtypes from ndx-pose, ndx-photometry, ndx-fiber-photometry #1665