SweepFormula: Add formula for spontaneous PSC detection (PSX)

[t-b]

• Sweep data in IC
• Filter data to reduce noise
• Deconvolute, see PA plot code
• FindLevel for finding the events
• User needs a persistent way to reject events via GUI. Event rejection is per sweep
• Visualize accepted/rejected events with colored circle

Output:

• Absolute time of events
• Amplitude (from filtered data, not from the deconvoluted data)
• Maybe more later

See also https://docs.google.com/document/d/1oCx1l-M0SApEcEsPPfIO5iWFk_TENieX9CZ9Tb5i0Io

Have a look how http://www.neuromatic.thinkrandom.com/ does it.

Example data is on the FTP in the folder 2022-03-08 PSC example data.

• Keep the reject/accept state machine readable for machine learning later on

[t-b]

Meeting with Travis today had the following outcome. The questions were my questions.

---

Q: Is everything important plotted? I.e. Do you need more info to decide on accepted/rejected?
A: Yes all plotted. Nothing additional for now.

Q: Is there a default (accepted/rejected/undecided) for each event?
A: start with accepted, no undecided

Q: How should we implement the UX for peak acceptance/rejection?
One could make clicking the marker toggle the state. I would also use a non-filled circle.
A: - Toggle state with Ctrl+Mouse click on event.

• Allow to use marquee with custom menu entries as well to change a range of events.
• Add keyboard control (request from @timjarsky) to do it without mouse and minimum key strokes. Add back button. Automatic jump to next.

Q: tau_d/tau_r are currently not calculated. Should they?
A: Not now.

Q: Does it make sense to plot the events on the deconvoluted data?
A: Yes makes sense.

Q: Should we plot the raw data close as well?
A: No, but plot filtered data, as already done, scale to visible, mark events in that as well

Additional plots to see how it evolves with the sweeps:

• Number of events per sweep (one for each accepted/rejected)
• Averaged amplitude of all accepted events per sweep

No export in NWBv2 data structures for now.

Tunable parameters:

• Deconvolution kernel
• find_crossings: thresh
• time to ignore at end of trace
• filter_sweep: low/high (optional)

Time range:

• Always ignore inserted TP

---

I said I come up with a first SF draft in around 2 weeks.

[t-b]

@timjarsky #MissingTravisMentioning

So I came up with a draft of the UI and the SF interface. See the attached experiment:
Graph2.zip

For convenience here is the graph:

The keyboard controls for acceptance/rejectance is partly already working. I did that so that we can get a feeling if this is going in the right direction.

I'm available for a chat today.

[t-b]

Comment from @MichaelHuth:

As for the special plot: Instead of looking for the last operation I think it is
better to use the meta data like SF_META_DATATYPE.
Or to introduce a new field for the wRef wave like SF_META_PLOTTYPE that a operation
can set.

[t-b]

@travis-open See #1313 (comment).

[t-b]

From https://aiephys.slack.com/archives/D0388H0EVDJ/p1666717915656869:

We can assume analysis/visualization of one HS at a time.