Updates to epilepsy metrics and plotting

[DowntonCrabby]

Describe the Issue
The epilepsy probability and interichtal events plot currently only uses the first 5 minutes of the motion corrected movie for its calculations. This should be run for the whole movie. Additionally the failure threshold is currently if the epilepsy probability metric is above 0%, but instead of failing the session, this actually triggers qcers (and usually a manager/scientist as well) to review the 2p movie and behavior movie to determine whether or not there was actual epileptic activity.

This process seems subjective and up for interpretation. Two additional questions arise: are we worried about aberrant brain activity or animal health? Depending on our concerns we might have different approaches to what causes data or a mouse to fail.

Clearer guidance is needed.

Expected behavior

• Calculations should be run on the entirety of the motion corrected movie and not the first 5 minutes.
• clear flag and failure thresholds should be established
• a protocol for manually reviewing the data should be established, with specific guidance on animal health & behavior, and separately aberrant brain activity

Scope
Interichtal events are usually related to a few specific cre lines but this is really run on every experiment and has a broad scope.

Additional context
In the future we should also investigate whether the presence of transient laser/scanner noise (issue #1 )on the Mesoscope impacts the probability calculation.

There could also be issues with the lowered frame rate per FOV on the mesoscope impacting this calculation.

This was originally a trello card: https://trello.com/c/xwCx8qMa/50-issues-with-epilepsy-plot

[samiamseid]

Following up with details from a recent slack thread on this topic:

Examples of a session that triggers the current epilepsy metric, but we suspect is not due to epilepsy: \allen\programs\mindscope\production\learning\prod0\specimen_1213623863\ophys_session_1221545894\ophys_experiment_1221714233\processed\1221714233_suite2p_motion_preview.webm
http://mouse-seeks/qc/1223267374
http://mouse-seeks/qc/1223027165

Marina suggests that we calibrate the metric for each reporter line due to differences in event magnitudes.

Literature on event magnitudes in interictal events https://www.eneuro.org/content/4/5/ENEURO.0207-17.2017

[matchings]

Another relevant paper:

https://www.sciencedirect.com/science/article/pii/S0896627316309114

[matchings]

@seanmcculloch has this been resolved in the new mouse QC system, or does it need to be included in the "need to haves"?