Skip to content

Outlier Detection #21

@aatamian

Description

@aatamian

Could we incorporate outlier detection within the wizards staff pipeline?

Modified Z-score on per-neuron summary statistics:
For each neuron from the df/f npy file, compute three metrics:

  • max df/f
  • mean df/f
  • standard deviation
    Then for each metric across all neurons, calculate this:
  1. The median of that metric across the population
  2. The MAD (mean absolute deviation) - the median of the absolute difference from the median
  3. A modified Z-score for each neuron: (0.6745*(value-median)/MAD

Any neuron within a modified Z-score above 3.5 (in absolute value) gets flagged.

Although it won't catch artifacts that happen to stay within normal df/f range, we can set up some sort of waveform template matching or frequency-domain analysis in the future.

Metadata

Metadata

Assignees

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions