EEG discriminant analyses of mild head trauma

Measures of EEG power spectral analyses were obtained from a total of 608 mild head trauma patients and 108 age-matched normal subjects. A training-set discriminant function was first developed from 264 mild head-injured patients and 83 age-matched controls yielding an overall discriminant classification accuracy of 94.8%. The first independent cross-validation of the discriminant function using 130 mild head trauma patients and 21 age-matched normals yielded a discriminant classification accuracy of 96.2% for the trauma patients and 90.5% for the normals. A second independent cross-validation of the discriminant function using 51 patients and measures of test-retest reliability from 93 patients yielded classification accuracies ranging between 77.8% and 92.3%. A third independent cross-validation of 70 mild head-injured patients tested at a different location with a different EEG computer system yielded a discriminant accuracy of 92.8%. The discriminating EEG power spectral analyses indicated 3 classes of neurophysiological variables which are attributable to mechanical head injury: (1) increased coherence and decreased phase in frontal and frontal-temporal regions; (2) decreased power differences between anterior and posterior cortical regions; and (3) reduced alpha power in posterior cortical regions.