The effect of a scalp reference signal on coherence measurements of intracranial electroencephalograms.

OBJECTIVE: To determine the effect of a scalp reference signal, such as that recorded from the mastoid or ear, on the coherence of referential intracranial electroencephalograms (EEGs).
METHODS: The relationship between reference signal power and magnitude squared coherence (MSC) was determined from the theoretical expression of the coherence of referential recordings, obtained under the assumption that the reference signal is not correlated with the signals being studied. The effect of a contaminated reference signal on the coherence of intracranial EEGs was determined by measuring the MSC of both a recording of background EEGs with a simulated contaminated reference signal and a contaminated recording of a seizure.
RESULTS: The MSC of referential intracranial EEGs is inflated due to the reference signal. This inflation is a function of the true MSC of the intracranial signals and the power of the reference and intracranial signals. The inflation is limited where reference signal power is smaller than the power of the intracranial signals; maximum inflation <0.1 when reference signal power=0.2xpower of intracranial EEGs and </=0.2 when reference signal power=0.5xpower of intracranial EEGs. A contaminated reference signal may have a considerable effect on MSC, however. The changes to the MSC spectrum that result from a contaminated reference primarily occur in an open-ended high-frequency band and may create an elevated plateau in this part of the MSC spectrum.
CONCLUSIONS: The findings presented here suggest a scalp reference signal is suitable, with careful monitoring of the reference signal, for coherence analysis of intracranial EEGs. A reference signal will have a limited effect on the coherence of intracranial EEGs except when it is contaminated. In the event the reference signal is contaminated it should be possible to detect this from the stereotyped features of the coherence spectrum.