quantifying FOV matching across sessions

[matchings]

Metric / Plot Impacted This is a proposal for a new metric, suggested by @jkim0731, to quantify the difference in z-plane location of a field of view across multiple sessions by comparing the average image of each session with the local z-stack and computing the difference in z, in microns, similar to how the z-drift metric is computed. This could be done for an average image taken from the beginning and end of the session to incorporate within session z-drift as well.

Update Motivation Currently FOV matching across sessions is done by eye on the rig and during container QC. A more quantitative metric would be valuable as a QC tool and potentially as an aid to operators during acquisition.

Requested Feature/Solution Application of this new metric to existing Visual Behavior dataset to evaluate how much across session shift is typically present. We can then calculate the expected % of cells lost across sessions due to the shift and set a new threshold on this metric and start applying it to new datasets.

Scope This would apply to all projects that use cell matching.

TBD: Please provide the following information as appropriate:

• data streams / files that are used
• formulas/calculations
• threshold for "flag" and "failure"
  • for plots: plot type, axis labels, axis scaling,

Screenshots or example plots

• current version of the plot
• sketchup/example of how updates to the plot would look

Additional context & analysis Add any other context or screenshots about the feature request here.

• links to jupyter notebooks of analysis and/or plotting examples

[matchings]

@jkim0731 has code to do this quantification, and it will need to be added to the Code Ocean QC process eventually (once container assignment logic and cell matching are put into place). @arielleleon will likely be the one to implement.