Spike detection is based on a very fast, online-capable method developed by Oliver Muthmann described here. Following detection, spatial spike locations are estimated, and redundant events removed. The results are stored in a flat binary file for fast access.
Assuming we have a created a Probe object called P, use:
H = HSDetection(Probe, to_localize, cutout_start, cutout_end, threshold,
maa=0, maxsl=12, minsl=3, ahpthr=0, out_file_name=out_file, save_all=False)Parameters:
- to_localize - If set to
False, spikes will only be detected, not localised. Note the data cannot be sorted then. - cutout_start - Number of frames to save backwards from spike peak.
- cutout_end - Number of frames to save forward from spike peak.
- threshold - Detection threshold, this is given in multiples of the estimated noise level. Note the variability measure is not a variance, but a percentile much better suited for the highly non-Gaussian noise in the data. As a rule of thumb, this can be set to around 10.
- maa - Minimal average amplitude of the event peak.
- maxsl - Dead time in frames after spike peak, used for further testing.
- minsl - Number of frames used for determining average spike amplitude.
- ahpthr - The signal should go below this value within maxsl-minsl frames
- out_file_name - Base file name (without extension) for the output file(s).
- save_all - If
True, store additional debug output.
Now we can read the result like so:
H.LoadDetected()