Evaluation of automatic quality assessment of auditory evoked potentials in clinical data

In a clinical study during cardiac surgery, the use of the (±) average for residual noise estimation was evaluated to assess the quality in the measured auditory evoked potential signals. The visual screening of the signals by three human experts was used as a reference. They categorized approximately 27% of the 683 AEP´s as “OK”, which could also be classified as such with a significant difference from other categories by the automatic (a) average method. The performance of the method was tested by combining the (±) average quality factor with the number of sweeps contributing to the average. The sensitivity for detecting AEPs in the “OK” category was only ~0.5, whereas the specificity was ~0.8 for correct classification of waveforms of insufficient quality