Reliable computer assisted classification of the EEG: EEG variants in index cases and their first degree relatives.

R.M. Dünki, G.B. Schmid, P. Scheidegger, H.H. Stassen, G. Bomben & P. Propping, American Journal of Medical Genetics 67: 1-8 (1996)
Keywords:
Discriminant analysis, Cluster analysis, Personal Channel Pattern, Heritability, Mendelian Interpretation

A method which optimizes on global properties of sample recordings is proposed for the definition of and the discrimination between EEG classes. The sample was drawn from students at the University of Heidelberg during 1974 to 1978 and consists of 15 healthy index cases clinically ascertained as belonging to the low voltage EEG group. In addition, the three clinically defined groups: diffuse (12 index cases), borderline (18 index cases) and monomorphous (18 index cases) have been included into the study as well as the first degree relatives of the index cases, thus providing a clinical classification into 4 groups. The proposed method provides an automatic and reliable classification algorithm using discriminant and cluster analysis. The relation between such an automatized classification and clinical classification schemes is investigated. In particular, the inheritance of of the low voltage EEG, the question on sex differences and the question of a simple Mendelian mechanism had been examined. The method of random splittings had been applied for discriminant and cluster analysis. Our findings can be summarized as follows: 1) Except for the monomorphous EEG group, the clinical classification shows rather marginal separation (discriminating performance 60 % to 75 %), while a new and more reliable grouping scheme improves the discriminating performance up to 87 % to 91 % . The latter scheme leads to the concept of personal channel pattern (PCP) and was compared to the clinical classification scheme by means of contingency tables; 2) Only a weak correlation between the clinically and PCP-based groups could be found (Cramer Index: 0.27). Accordingly, we continued to investigate the extent to which the proposed EEG classification scheme can nevertheless explain the genetic mechanisms apparently involved in the low voltage EEG. We thus considered the role of sex differences manifest in our proposed new grouping scheme. 3) Males occurred more frequently in the new group 3 and females more frequently in the new group 1. In this regard, a much better correlation of the new groups between mothers and children than between fathers and children was observed. 4) With help of our new PCP-scheme, we have been able to reproduce a simple two gene Mendelian scheme to explain inheritance of the clinical low voltage EEG group. In this PCP-based scheme, the low voltage property does not occur when dominance of a certain gene (called gene A) is absent.