A new quantitative measure for monitoring somatosensory evoked potentials

This paper describes the development and testing of a computer algorithm to automate the process of peak identification and somatosensory evoked potential (SSEP) grading. We tested the accuracy of computerized peak detection and evaluated grading schemes using a test set of 60 SSEPs ranked from worst to best by the programmer (RJM) and a blinded grader (PO). The computer algorithm recognized 95% of peaks identified by visual inspection. Twelve percent of peaks identified by the computer were noise. Summed peak to peak amplitude gave the most accurate ranking of SSEPs. Rank correlation between computer and blinded and unblinded expert grading was r = .82 for PO, r = .92 for RJM, p < .0001 for both. Computer and manually summed amplitudes were highly correlated (Pearson r = .98, p < .0001). Correlation between the 2 expert graders was .86, p < .0001. Computer graded SSEPs were significantly related to clinical outcome at 3 months, p < .0001. Automatic grading of SSEPs using summed peak to peak amplitude is highly correlated with expert grading. The measure is objective, continuous, and well suited to statistical analysis.