Figure 3. Isolation of currents containing contributions from an oscillatory LFP of interest.

Application of the procedure to experimental and model LFPs are presented side by side to facilitate the biophysical interpretation of signal transformations (left and right columns, respectively). A shows sample traces of raw LFPs at different domains along the pyramidal cell axis: colored stars mark recordings in the st. pyr, rad. and l-m, respectively. Model LFPs were high-pass filtered (>0.1 Hz) to reproduce AC-coupling of experimental recordings, and the distribution of the inputs was simpler than in the real CA1 to better visualize the changes in oscillatory input. CSD analysis of LFPs produced a complex spatiotemporal mixture of current sinks and sources. Few or no domains of active synaptic sites (cf. Figure 1) were detected. The amplified segments below show that individual gamma waves in the st. rad. may be matched by either sources or sinks at the active Schaffer domain. B shows the separation of the pathway-specific contributions by ICA applied to raw LFP profiles. Three different generators (G1-G3) were obtained, each of which defined by the characteristic spatial distribution (weight at each electrode) and temporal activation specific to the period analyzed. The respective spatial distributions are shown in the middle. Those obtained for model LFPs tightly reproduced the distribution of synaptically activated compartments, and the temporal sequence of inputs was accurate. Following reconstruction of pathway-specific LFP profiles, the application of CSD analysis rendered a spatiotemporal map of sources and sinks in which stable reversal sites were observed. The model confirmed that these corresponded to the macroscopic boundaries of active and passive domains of the synaptic input. However, note that both sinks and sources still appeared at the synaptic site. These temporal patterns do not allow us to determine whether sinks or sources at each domain are real or spurious.