Signal verification of middle latency auditory evoked potentials by automated detection of the brainstem response.

BACKGROUND

The midlatency components of auditory evoked potentials (AEPs) are gradually suppressed with increasing concentrations of anesthetics. Thus, they have been proposed as a monitor of anesthetic depth. However, undetected malfunction or disconnection of headphones and undetected hearing loss also result in suppressed midlatency AEPs that in turn may be misinterpreted as signs of deep anesthesia. As the brainstem component of the AEP is minimally influenced by anesthetics, its presence or absence can be used to verify that the recorded signal is a true AEP rather than an artifact. In this study, an online-capable procedure for detection of the brainstem component of the AEP was developed.

METHODS

One hundred and ninety perioperatively recorded AEPs (binaural stimuli, 500 sweeps) were selected from a database with electroencephalographic and concomitant AEP stimulus information. Identical electroencephalogram regions were used to produce nonstimulus synchronized averaged signals (500 sweeps, "non-AEP"). The 190 AEPs and 190 "non-AEPs" were used to develop a detector of the brainstem component of AEPs. AEPs and "non-AEPs" were wavelet transformed (discrete wavelet decomposition, biorthogonal 2.2 mother-wavelet), and the coefficient with the best separation of the two classes of signals was selected. Receiver operating characteristic curve analysis was performed to determine the optimum threshold value for this coefficient.

RESULTS

The third coefficient of the third level was selected. In AEP signals, retransform of this coefficient produces a peak that resembles peak V of the brainstem response. The developed detector of the brainstem component of AEP had a sensitivity of 97.90% and a specificity of 99.48%.

CONCLUSIONS

This detector of the AEP brainstem component can be used to verify that the signal reflects the response to an auditory stimulus. An alternative approach, used in the Danmeter AEP monitor, is based on the signal-to-noise ratio of the midlatency components of the AEP. Because the midlatency components of AEP are suppressed by anesthesia, a false alarm "low AEP/no AEP" is generated during deep anesthesia. This, in turn, may suggest disconnection of headphones or technical problems whenever anesthesia is deep. This disadvantage has been overcome by our detector, which is based on the identification of the brainstem component of AEP.