Ignore symmetric deconvolution timeseries

[timjarsky]

Analysis of current clamp data with PSX results in deconvolution traces that surpass reasonable detection thresholds but are not associated with actual synaptic events. A feature of these false positives is that the deconvolved traces are symmetric (see image). Would it be possible to filter out these false positives by looking for symmetry in the deconvolved trace? Perhaps a search for an equal opposite sign threshold within a time window would be sufficient.

Open question: should this be an operation where the user can turn it off and adjust other related parameters?

blue box: false positive with symmetric deconvolved trace green box: miniature EPSP (excitatory post-synaptic potential)

[timjarsky]

Another way to filter false positives like this would be to correlate the peak amplitudes of the deconvolved events with the event amplitudes and remove the outliers.

This wouldn't work well when the number of events is small