ANALYTICAL METHODS TO DIFFERENTIATE SIMILAR ELECTROENCEPHALOGRAPHIC SPECTRA - NEURAL-NETWORK AND DISCRIMINANT-ANALYSIS

Differences in electroencephalographic (EEG) power spectra obtained un der similar, but not identical, conditions may be difficult to discern using standard techniques. Statistical analysis may not be useful bec ause of the large number of comparisons necessary. Visual recognition of differences also may be difficult. A new technique, neural network analysis, has been used successfully in other problems of pattern reco gnition and classification. We examined a number of methods of classif ying similar EEG data: standard statistical analysis (analysis of vari ance), visual recognition, discriminant analysis, and neural network a nalysis. Twenty-nine volunteers received either thiopental (n = 9), mi dazolam (n = 10), or propofol (n = 10) in sedative doses in 3 differen t studies. These drugs produced very similar changes in the EEG power spectra. Except for beta2 power during thiopental infusion, difference s between drugs could not be detected using analysis of variance. Visu al categorization was correct in 72% of the baseline EEGs, 70% of thio pental EEGs, 27% of propofol EEGs, and 46% of midazolam EEGs. A classi fication neural network (Learning Vector Quantization network) contain ing a Kohonen hidden layer was able to successfully classify 57 of 58 EEG samples (of 4 minutes' duration). Discriminant analysis had a simi lar rate of success. This level of performance was achieved by dividin g the EEG power spectrum from 1 to 30 Hz into 15 2-Hz bandwidths. When the EEG power spectrum was divided into the ''classical'' frequency b andwidths (alpha, beta1, beta2, theta, delta), both neural network and discriminant analysis performance deteriorated. By training the netwo rk using only certain inputs we were able to identify drug-specific ba ndwidths that seemed to be important in correct classification. We con clude that propofol, thiopental, and midazolam produce different effec ts on the EEG and that both neural network and discriminant analysis a re useful in identifying these differences. We also conclude that EEG spectra should be analyzed without using classical EEG bands (alpha, b eta, etc.). Additionally, neural networks can be used to identify freq uency bands that are ''important'' in specific drug effects on the EEG . Once a classification algorithm is obtained using either a neural ne twork or discriminant analysis, it could be used as an on-line monitor to recognize drug-specific EEG patterns.