Enhancement of signal-to-noise ratio of peroneal somatosensory evoked potential using independent component analysis and time-frequency template

Cortical somatosensory evoked potential (SSEP), recorded by electroencephalography (EEG) system, has been used to assist the diagnosis of certain neuropathologic states. The evoked potential induced from patients with impaired motor functions, however, is potentially much smaller than the background EEG activities and exhibits longer latency and lower amplitude than the normal SSEP. In this study, we aimed to recover the SSEP induced by right and left peroneal nerve stimulations from smearing EEG recordings using independent component analysis (ICA) in conjunction with the temporal SSEP template or the proposed time-frequency SSEP template. Since the SSEP is inevitably contaminated by artifacts and environmental noise, we employed ICA to decompose the EEG into a set of independent sources from which the SSEP-related features were automatically selected by the proposed template for reconstruction. Our results from three normal subjects and four stroke patients demonstrated that the proposed method could remarkably suppress the artifacts and effectively reconstruct SSEP waveforms in comparison with the conventional averaging method.