A measure of connectivity in the presence of crosstalk

Coherence is a widely used measure to investigate interaction between electrophysiological signals obtained from EEG or MEG. However spurious coherence between locations can be introduced through cross-talk or linear mixing between signals that occurs as a result of diffuse lead-field sensitivity (in sensor data) or limited spatial resolution (in cortical current density maps). Measures including the imaginary part of the coherence (IC), phase lag index (PLI) and Lagged coherence (LC) have all been proposed to overcome this problem. Each of these metrics use the fact that cross-talk will produce instantaneous or zero phase-lag interactions between signals. By constructing measures such as these, all of which have zero value in the case of instantaneous mixing only, we can reduce sensitivity to cross-talk. However, none of these measures considers the effect of interference from external sources. In this paper we first investigate the relative performance of IC, PLI and LC. We then propose and evaluate a novel measure, partial lagged coherence (PLC), which is more robust to cross-talk in the presence of interfering signals.