Detection of transient single-sweep somatosensory evoked potential changes via principal components analysis of the Autoregressive-with-exogenous-input parameters

A method for detecting transient or rapidly occurring changes in evoked potentials is presented here. The recorded electroencephalogram is modelled as an AutoRegressive process with eXogenous (ARX) input. This model is used to estimate single-sweep somatosensory evoked potentials (SEPs). Principal components analysis is then applied to the ARX coefficients, reducing to six the number of parameters needed to account 95% of the SEP variability between sweeps. Application of this method to single-sweep SEP analysis in human impact acceleration experiments provides a critical improvement in temporal resolution over ensemble average estimation. This method makes investigation of transient SEP changes associated with impact acceleration possible.