Unit curation based on amount of spikes

[b-grimaud]

Is there any way to filter out units based on the amount of spikes they contain ?
I've looked through the post-processing and quality metrics modules and couldn't find any clear way to do so, and I'm not sure if it might be embedded into another extension.

Data from HD-MEAs can yield a few thousands prospective units after sorting, and a simple threshold on the amount of spikes (something like > 10) can remove a good chunk of false units. This in turn helps with computation time down the line.

[samuelgarcia]

Both Sorting and SortingAnalyzer has a remove_units() method.

Quality metrics has a "num_spikes" columns that can be used to select theses.

[b-grimaud]

Thanks !

The issue I had with this was that some parent extensions of quality_metrics, especially waveforms, can require a lot of time and memory (100+ GB) when computing on many (possibly erroneous) spikes.

So far I found that the fastest way is simply to iterate over each unit of the sorter, test the length of the corresponding spike train and remove the smaller ones, then create the analyzer.