Thresholds for cluster quality


I’m looking at the cluster quality metrics as seen here:

sl = SpikeSortingLoader(pid=probe_ids[0], one=one, atlas=ba)
spikes, clusters, channels = sl.load_spike_sorting()
clusters_labeled = SpikeSortingLoader.merge_clusters(spikes, clusters, channels)

I’m wondering whether there is any standard/good cutoff to exclude clusters for being poor quality (e.g. whether a cluster is scored as 0.66 vs. 0.33)?

In addition, is there any way to know whether a cluster is a multiunit that shouldn’t be separated?