There are 2 things that would make quality metrics more useful:
make sure that we record thresholdable values in the final dataset, so that the treshold can be readjusted. This is trivial for amplitudes, but for the sliding refractory metric we may want to record the lowest confidence at which a given unit passes
encode the metric result in a integer where each bits encodes whether or not a QC passes or fail. In this case it is easier to have failures set to 1 so that an overall 0 value means all qcs pass, regardless of the number of QCs
Also there is no reason to have this as a separate task. This data should be part of the spike sorter output, especially since the environments are fully compatible.
There are 2 things that would make quality metrics more useful:
Also there is no reason to have this as a separate task. This data should be part of the spike sorter output, especially since the environments are fully compatible.