Multivariate objective response detectors (MORD):statistical tools formultichannel EEG analysis during rhythmic stimulation
Abstract The presence of cerebral evoked responses can be tested by using objectiveresponse detectors. They are statistical tests that provide a threshold above which responses can beassumed to have occurred. The detection power depends on the signal-to-noise ratio (SNR) of theresponse and the amount of data available. However, the correlation within the background noisecould also affect the power of such detectors. For a fixed SNR, the detection can only be improvedat the expense of using a longer stretch of signal. This can constitute a limitation, for instance,in monitored surgeries. Alternatively, multivariate objective response detection (MORD) could beused. This work applies two MORD techniques (multiple coherence and multiple component synchronymeasure) to EEG data collected during intermittent photic stimulation. They were evaluatedthroughout Monte Carlo simulations, which also allowed verifying that correlation in the backgroundreduces the detection rate. Considering the N EEG derivations as close as possible to the primaryvisual cortex, if N = 4, 6 or 8, multiple coherence leads to a statistically significant higherdetection rate in comparison with multiple component synchrony measure. With the former, the bestperformance was obtained with six signals (O1, O2, T5, T6, P3 and P4).
Keywords Acoustic Stimulation; Electroencephalography
Annals of Biomedical Engineering
0090-6964, Volume 35, Issue 3, 2007, Pages 3-452
