search

AllenInstitute / brain_observatory_qc

Other
2 stars 0 forks source link

Dev-QC: Non-zero event probability should prompt a failure #329

Closed arielleleon closed 9 months ago

arielleleon commented 10 months ago

The "Event Probability (%)" column in the Interictal Events report module, when non-zero, does not automatically trigger a failure and it should

DowntonCrabby commented 10 months ago

Looks like the thresholds were missed in the new database but it's been updated now.

A FLAG is now triggered when the event probability is not 0- and then the motion corrected video MUST be reviewed before and the project lead notified before a fail is added- that is what was discussed in one of the ops meeting. Maybe @matchings can weigh in on if that's still the strategy we want to take rather than an unreviewed failed.

image

matchings commented 10 months ago

@DowntonCrabby i agree that its a good idea to have a project lead review these cases, as there are a lot of weird things that can happen that we may not have encountered yet, and/or things that are totally biologically normal that could trigger this metric (like the cre line being one for fast spiking inhibitory cells). This isn't to say its not a good metric, I actually think its one of the most useful ones we have given the range of situations it can pick up. But given that, it does warrant a closer look by both operators and science leads.