Principle for modelling and subtracting the LFP.

(A) Rostrocaudal localization of the laminar electrode array used to record the local field potential (LFP) in the mouse hippocampus. (B) Localization of the laminar electrode array in the coronal plane of the mouse brain (lateromedial and dorsoventral localization). (C) The panel shows the time windows used to model the LFP. The dashed red line is a tricube window (here: 100 ms wide) used to get a local estimate around the spike in the center (0 ms). The dashed green line is flipped tricube windows used to remove the influence of the spike waveform from the fitted LFP model. Three spikes are edited out. The solid red line is the weight function actually used to estimate the regression model, computed as the product of the dashed red and green lines. The corresponding LFP signal is shown in blue. The dashed black line is the DC level. (D) Depth profile of temporally weighted LFP traces recorded from the mouse hippocampus. A microscope image of the tip of the laminar electrode array is shown in the background. The 16 light spots in the center of the probe are the recording sites. The electrode array can be seen to cover the full depth profile trough the hippocampal layers from CA1 to DG. Recording from a set of reference channels in a laminar electrode can give an estimate of the local field potential on a different channel with spikes, thus allowing the estimated LFP to be subtracted in order to recover the spike waveforms from the LFP. The recovered waveforms are expressed as the prediction errors. The illustrations of the mouse brain (A, B and D) are adapted from Paxinos and Franklin [56].