BIDS for systems neuroscience

[just-meng]

Hi all,

@yarikoptic asked me to review a few issues (#1632, #2137, #162) and I thought it would be nice to tackle the challenge more systematically in a separate issue. Since I know very little about the existing BIDS, I don't aim to solve anything, but simply to contribute a point of view from someone who works in experimental neuroscience at systems level (right between the fMRI/EEG and molecular stuff, mostly animal research). It is a rapidly growing field within neuroscience, in particular, due to tool development (genetics, optics, probes etc.). There is no common standard yet in systems neuroscience as far as I know. So a lot to gain here!

Rationale

The current microscopy suffix table mixes optical physics mechanisms with data modalities and experimental context.

• Terms like FLUO and 1P are non-discriminative in neuroscience (almost everything is fluorescence microscopy, and everything that is not 2-photon/3-photon is 1P).
• Distinction between 2P and 3P is overgranular, multi-photon (MP) is fully sufficient here. Skip the E, since almost everything is excitation of fluorecence proteins.
• Binary categories such as ephys and ophys is quite misleading. Strictly speaking, only voltage imaging is optical physiology. Every other method that does not record the membrane potential is not physiology. Calcium imaging could be considered ophys, but is not really commonly referred to as such. I bet most people from my field wouldn't know what is meant with ophys. I'd rather suggest electrophysiology and imaging/microscopy. It's a hard call between imaging and microscopy. I am aware of imaging data in my field that are non-optical (but acoustoptical, electron-microscopy, X-Ray, ultrasound etc.) or not with a microscopic resolution (e.g. mesoscopic). Choose imaging for all pixel-absed spatial acquisition, and choose microscopy if you want to include mainly optical microscopic data which IS the most abundant data type after ephys.
• What is mostly missing, but matters a lot in systems neuroscience is whether the prep is in vivo (awake, anesthetized, freely-moving), ex vivo (fixed tissue, cleared brain slices, acute slice etc.), or in vitro (cell culture etc.). Although this axis is important for data sharing/discovery, it does not really matter for the data type. Although in vivo data are often accompanied by behavioural data and/or perturbation data.

Based on these considerations, I suggest organizing by scientifically meaningful axes:

• What is measured (biological signal)
• How it is acquired (acquisition technology)
• In what context (biological preparation)

New Metadata Axes

Field	Description	Examples
measured_signal	Biological signal being recorded	calcium, voltage, hemodynamic, structural, metabolic, other
acquisition_method	Imaging/acquisition mechanism	see controlled vocab below
preparation_context	Biological experimental context	in_vivo_awake_head_fixed, ex_vivo_acute_slice, fixed_cleared, etc

These axes are orthogonal and avoid ambiguous categorization. They are not exhaustive.

Controlled Vocabulary: acquisition_tech

Category	Term	Notes
Multiphoton scanning	multiphoton_scanning
Widefield epifluorescence	widefield_epifluorescence	Replaces FLUO
Mesoscopic widefield	widefield_cortexscale	Cortex-wide imaging
Miniscope / endoscopic	miniscope_endoscopic	Head-mounted 1P/2P
Light-sheet	light_sheet	SPIM, mesoSPIM
Confocal	confocal	Includes spinning disk
Super-resolution	super_resolution
Electron microscopy	electron_microscopy
X-ray tomography	xray_tomography	XPCT / µCT

Cheers!

[yarikoptic]

attn @CodyCBakerPhD @bendichter @bids-standard/bep031 and might be of interest to some @bids-standard/bep032

[CodyCBakerPhD]

roping in @alessandratrapani as someone who can comment with much more authority on the matters discussed

[saskiad]

I wouldn't separate mesoscopic widefield and would use multiphoton scanning instead.
Unless you are using mesoscopic differently than I am? We routinely use a the Thorlabs Mesoscope to image with the same spatial resolution as our 2P microscopes.

Otherwise I agree with this assessment.

[just-meng]

Thanks — great point. To clarify, I did not mean the Thorlabs Mesoscope 2-photon system.
When I say mesoscale widefield, I'm referring to 1-photon large FOV epifluorescence imaging (across cm-scale windows).
Multiphoton Mesoscope platforms should remain under multiphoton_scanning. I agree!
Perhaps is better to rename this category into widefield_cortexscale to avoid confusion.