Characterization of neurologic injury using novel morphological analysis of Somatosensory Evoked Potentials

This paper describes an innovative, easy-to-interpret, clinically translatable tool for analysis of Somatosensory Evoked Potentials (SSEPs). Unlike traditional analysis, which involves peak-to-peak amplitude and latency calculation, this method, phase space analysis, analyzes the overall morphology of the SSEP, and includes greater information. The SSEP is plotted in phase space (ẋ vs. x), which leads to an approximately spiral curve. The area swept out by this curve is termed the Phase Space Area (PSA). As PSA calculation involves numerical differentiation, we present a comparison of two different approaches to combat noise amplification: finite-window smoothing, and total variation regularization (TVR) of the numerical derivative. These methods are applied to simulated SSEPs. The efficacy of these methods in performing noise-reduction is assessed and compared with ensemble averaging. While TVR gives a reasonably robust approximation of the derivative, Gaussian smoothing of the derivative offers the best trade-off between the number of signal sweeps required to be averaged, close approximation of the SSEP derivative, and optimal estimation of the PSA. We validate this method by analyzing non-characteristic SSEPs that have indistinguishable peaks as is frequently seen in cases of underlying neurologic injury such as hypoxic-ischemic encephalopathy.