Re-evaluate intensity drift (%) metric

[jkim0731]

Describe the Issue

• Clear intensity drift (probably due to laser instability) not captured by the metric.
• Current intensity drift metric flags for fine experiments.

Screenshots

https://alleninstitute-my.sharepoint.com/:p:/g/personal/jinho_kim_alleninstitute_org/EeRvwo3hEB9Lro1fVVH1tJIBbYvum2CtDe1yG6SMlnAZGA?e=lmDFor

Scope

• Intensity drift (along with low values) were seen in the recent MESO.2 before the laser change (10/23/2023).
• Since laser change it is fine.
• This needs to evaluated better to detect laser instability in the future (or for MESO.1).
• False flag affects ALL the experiments.

[jkim0731]

• Control language tag
• Preventive approach of checking laser stability (e.g., every 2 months, quarterly, etc., imaging fixed sample such as fluorescence beads or pollen grain)

[seanmcculloch]

For now, Operator's should use the "other" tag to flag the fluorescence plot. During our review of "Other" tags, we can create meaningful controlled language tags around various failures, and potentially update the intensity drift metric (or another metric) to be more comprehensive

[DowntonCrabby]

the % change currently checks the first portion of the timeseries to the last portion of the timeseries- so you're correct in that it won't catch any intermediate drastic changes.

Maybe we can make some automated metrics that would be able to catch this- like a sliding window , or standard deviation. What do you think @jkim0731

[samiamseid]

connected to following issues:
#317
#188
#67