Best Practice for EMG data?

[pauladkisson]

The Best Practices document states:

ElectricalSeries are reserved for neural data. An ElectricalSeries holds signal from electrodes positioned in or around the brain that are monitoring neural activity, and only those electrodes should be in the ElectrodeTable.

For non-neural electrodes that still may store and report raw values in Volts, simply use a general TimeSeries object with units set to “Volts”.

But, EMG naturally lends itself to ElectricalSeries since it has a location (muscle group), recording device, filtering parameters, A2D conversion and offset, and potentially relevant information about the electrode contacts.

So, I think we should amend the best practices to include all bioelectric signals in ElectricalSeries. Or else, we should have a bit more guidance on these edge cases.

Also see #327 which recommends ElectricalSeries even in the case of non-biological auxiliary channels (ttls, etc.).

[pauladkisson]

Summary of meeting 4/15

• Oliver and Ryan think in general it is worth the trouble to implement an extension just for renaming/clarity
• w/o different types, a dandi user couldn't know whether an electrical series is LFP, ecog, eeg, emg, etc.
• based on ndx-microscopy, we might want to just add an enum attribute that specifies the signal type by text, maybe on the ElectricalSeries on the data

Overall proposal: Add an attribute to ElectricalSeries.data: 'signal_type' that would be enumerated text: "eeg", "ecg", "emg", "lfp", etc.

[stephprince]

To add some information about the discussion in our sync meeting to @pauladkisson's nice summary above. The main topic of concern that was brought up was that if we allow ElectricalSeries to contain any bioelectric signals we want to make it clear to a user what type of data is contained (e.g. EMG vs. EEG vs. ECG, etc.).

There were two main suggestions on how/where to store this information:

1. In the ElectrodesGroup of the ElectricalSeries.electrodes
2. As an attribute to the ElectricalSeries.data

The current ndx-microscopy extension distinguishes data types based on the excitation source in ndx-ophys-devices, but that case is a little different since it is the same signal acquired with different devices, while the EMG vs. ECG vs. EEG distinction can be similar devices (e.g. a microwire) but recording different types of signals.

The conclusion was that for ease of data reuse and where users would expect to find this information, adding an enumerated signal_type attribute to ElectricalSeries.data may be the best approach.

[h-mayorquin]

Meanwhile, the docs should be amended right?

I think the docs should be something like "if you have localization and information about the electrodes it is better to add this data as an ElectricalSeries so the data is properly linked. Add an informative name an description to indicate the type of this signal. Otherwise, and to avoid adding non-meantingful data just add it as TimeSeries".

[h-mayorquin]

To add more to the documentation discussion:

1. The current documentation for ElectricalSeries states that the neuroda type is meant extracellular data only:

https://github.com/NeurodataWithoutBorders/nwb-schema/blob/ade50ef33446beb3c7df4c6f1072ae0e821b5115/core/nwb.ecephys.yaml#L2-L8

2. In the intracellular case we handle things differently. There is an Intracellular Electrode and we don't use the electrodes table.

3. Other two cases two consider are ECoG and EOG (Electrooculography)

4. The newly added ElectrodeTable schema makes the same assumption about extracellular data. The documentation is ridden with terms that are very specific to extracellular electrophysiology: "brain region", "probe geometry", "atlases", "sterotatical coordinates", etc that don't really make sense for surface electrodes (although as @pauladkisson noted, some of them can be used for needle electrodes).

5. The electrode table needs a compulsory ElectrodeGroup and at the moment that was designed and meant to be used in extracellular electrophysiology. See this discussion here: Improve documentation of ElectrodeGroup nwb-schema#631

[h-mayorquin]

Thinking about this, I think that is too much of a stretch to use the ElectricalSeries for most cases of biopotential signals. I opened another issue in the nwb-schema to discuss this further:

NeurodataWithoutBorders/nwb-schema#651