| A chi-square minimization criterion for optimal modeling of event-related potentials (ERPs) is formulated, and two brain signal processors are derived: (i) a topographic matched filter (TMF) for detecting and measuring ERP signals on single sweeps, and (ii) a latency jitter deconvolution (LJD) method for extracting the waveforms of overlapping ERP components. The TMF processor utilizes signal and noise estimates across multiple scalp locations to measure latencies and amplitudes of detected signals. The method is statistically calibrated to limit the likelihood of a type I error (false detection of an ERP signal), to estimate the likelihood of a type II error (failure to detect an ERP signal), to estimate measurement errors, and to test for actual latency jitter and amplitude variability. The LJD processor generalizes ordinary time-locked signal averaging for multi-component ERP data by feeding on sweep-by-sweep latency and amplitude information provided by the TMF processor. These processors are applied to a sample of omitted visual stimulus data which contains emitted N200 and P300 components. Two studies are made. The first study assesses the TMF performance increments attributable to utilization of multi-electrode and background EEG spectrum information. It is concluded that multi-electrode information improves detection, estimation of latency, and sensitivity to latency jitter, and that background EEG spectrum information improves amplitude, latency, and waveform shape estimates. The second study applies the LJD processor to separate the overlapping N200 and P300 components. The waveform decompositions reveal allocations of signal energy not evident in the ordinary time-locked average. For example, positive deflections in the average at the Cz and Fz electrodes both appear to reflect the P300 component; however, the decomposition reveals that the Fz deflection is a polarity-reversed manifestation of the N200 component. It is hoped that these developments in neuroelectric signal processing will help to bring cognitive psychology and biophysical neuroscience closer together in the crucible of ERP research. |
