EDHMM classification produces improved agreement between simultaneously recorded LFP and MP signals.

A) Example of an LFP (blue trace) recorded simultaneously with the MP (black trace) of a nearby cortical neuron. The corresponding LFP (gray) and MP (red) state sequences inferred from the EDHMM are overlaid. B) The instantaneous probability of detecting false DOWN states is plotted against that for false UP states for the EDHMM method as well as the SMM- and Np-TC methods. The mean and SEM are indicated by the colored crosses. C) The probability of a missed LFP state, relative to the MP state sequence, is plotted against the probability of detecting an extra LFP state. D) Box plots illustrating the changes in e_i relative to the EDHMM algorithm for the following decoding algorithms: HMM; fixed-mean EDHMM (fm-EDHMM); static mixture model threshold-crossing (SMM-TC); and nonparametric threshold-crossing (Np-TC). E) Same as D for e_s.