Duplicate units with opposite polarity

[oaaij-gnahz]

Hi,

I was using SpikeInterface to spike-sort some 32-channel ephys data (MountainSort4). Somehow this phenomenon keeps occurring to me (see attached image): two likely duplicate units with opposite polarity but otherwise almost exactly the same template waveform. I wonder if it has happened to anybody before? Could it be electrical noise or "real" units?

I filtered the raw data with 300-6000Hz BPF (other settings are SpikeInterface default); used the median of all channels to re-reference; I did not whiten the data (whitening gave me other problems such as way too many units); MountainSort4 was configured to detect both positive and negative spikes with clip_size=60 and detect_interval=30 (recording was 30kHz). Are there something obvious that I was doing wrong?

Thanks for the discussion!

[alejoe91]

Hi @oaaij-gnahz

Great that you're using sorting view! Can I ask you why you use both positive and negative? I would use only negative since you know that the main spike is negative. Can you try that?

[oaaij-gnahz]

Thanks for replying so fast! I use both positive and negative because we previously did see some spikes that are positive. Also this paper's simulation result suggests that action potentials may look positive during transmission through axon, etc. (Fig 1 of that paper). I should definitely test using only negative, though.