Skip to main content
. 2020 Dec 1;11:6138. doi: 10.1038/s41467-020-19969-2

Fig. 2. Interictal spike identification and basic characterization.

Fig. 2

a Classifier schematic. Left: An Online Kernelized Perceptron was trained on the LFP magnitude spectrum where the window size was locked to each imaging frame. Right: Examples of classified interictal spike (above) and non-spike (below) highlighted in green, with expanded view (below left) and magnitude spectrum (below right). b Average IED magnitude spectrum. Average magnitude spectrum of n = 459 events from one subject, with 1/f power correction applied (data presented as mean value +/− 95% CI). A peak frequency occurs in the 250–400 Hz range, exaggerated by the 1/f correction, reflecting the timescale of positive and negative deflections (on the order of 10 ms in an event lasting ~50 ms), which may be polyphasic for some IEDs. c Summary of detected IEDs rate across mice (n = 5 mice). Left: Average IED rate, mean 1.13 IED/s, standard deviation 0.53 IED/s. Each point corresponds to a mouse (average of 3–5 sessions per mouse). Right: Total IED count per 10-min recording session, mean count = 680 IEDs/session, standard deviation = 319 IEDs/session (each point corresponds to a single session, each color corresponds to a different mouse). Error bars: +/− 95% CI in the mean, calculated from 1000× bootstrap sampling on recording sessions with replacement. d Spike-triggered PSTHs showing averaged response to IED by cell type, pooled across five animals (n = 239 abGCs, red; 1736 mGCs blue). The majority of GCs are not modulated by individual spikes on average (data presented as mean value +/− 95% CI).