Abnormally and spike events - Detect the abnormality in the system, given spontaneous data.

[skim0119]

No description provided.

[jihugo]

@skim0119 Here's my approach that I'd like to check with you before I go deeper into this:

1. Generate PCA cutouts for spontaneous recording
2. For each component, have the user label spontaneous spikes into ['neuronal', 'false', 'mixed']

• User isn't required to do label components for every single channel
• We can possibly make algorithms to suggest the user which channel to consider labeling

3. Train the model
4. For the experimental recordings:

• Catch the channels with higher uncertainties. These may indicate some new spike shape.
• (Possibly) Filter out channels with same ratio of neuronal to false spikes as spontaneous
• Generate new spiketrains with the false spikes removed

[skim0119]

@jihugo Sounds reasonable to me! 👍