Open bendichter opened 4 years ago
Hi,
I will be happy to help and provide more information for one-photon and two-photon imaging to make sure collected data makes datasets reproducible and promotes best practices in terms of appropriate experiment documentation.
Before going into low-level details, I would suggest a short discussion or just introducing me how do you currently see generic microscope recordings on the high-level.
From the diagrams found in: https://neurodatawithoutborders.github.io/matnwb/tutorials/html/ophys.html
It looks to me the key class currently is ImagingPlane
and perhaps TwoPhotonSeries
which of course does not cover any one-photon imaging which is really a widespread technique, probably with more users than two-photon these days.
We can logically break apart objects related to the apparatus, particular imaging session, surgical preparation of the mouse on the higher level, and then properly define each of those subcategories.
A few more comments on Ben's summary:
Objective lens:
manufacturer: str
magnification: float
numerical aperture: float between 0 and ~2, high-end objectives for two-photon have NA=1.1, the values can get much larger for in-vitro imaging with confocal microscopes (formally NA=n x sin(theta)
, and n
index of refraction n>=1
Standard 2PMicroscope
can be just broken down into objective lens (replaceable), laser model (may vary independently from microscope body), filters (set for individual measurements) and optionally PMTs (they are occasionally upgraded).
Custom 2p microscopes have just a few more components to be reported (and they always are there). I can comment on it later.
This is currently being addressed and tested through NWB extensions:
Radek (@chrapkiewicz) is looking at NWB for help sharing some upcoming optical data, and has some requests regarding device metadata. Here are the comments regarding 2p. (Radek, please correct and expand as necessary)
excitation power in Watts is one of the most important pieces of metadata and is currently missing. Normal range is 20-100 mW and upper end is 200 mW.
ImagingPlane
object.imaging_depth is another important parameter.
No one is using lumens. People record the power they apply, but never the power they collect. We should not require users to express their data in lumens.
TwoPhotonSeries
.The two most important things to document are the objective lens and the laser. For Objective lens, the most important features are:
For laser, the most important features are:
For the
OpticalChannel
associated with anImagingPlane
, there should be a linked dichroic filter. This would have features:There is a general rule that Radek recommended for devices that would apply to both the laser, objective lens, and the diachroic filter: The essential metadata would be fully captured if the user supplied the part model number. If the model number is given, the other data can be looked up and that should be counted as sufficient. I think this could be a general rule across modalities so I am going to make it a separate issue (#461), but I thought it was worth noting here.
I'm not sure how to reconcile the metadata requirements with the current data types.
ImagingPlane
requires a genericDevice
type. Maybe the most straightforward thing would be to define a2PMicroscope
type that contains aLens
andLaser
(both of which are alsoDevice
s), and then encourage users to use a2PMicroscope
instead of a genericDevice
.