A theory of laboratory data-based diagnosis and results of its application are reported. The diagnosis was made by linear discriminant analysis and matrix analysis, and the two diagnostic modalities were tested in healthy subjects and patients with 10 different diseases of the liver and the biliary tract. The optimal number of test items to be incorporated in the discriminant for screening was 2 to 3 in liver parenchymal diseases but tended to be 4 to 5 or more in obstructive or space-occupying diseases. A newly designed matrix discrimination was applied to this evaluation. By this method, all healthy individuals could be discriminated from patients. A mean percentage of 97.5% (92-100%) of patients with liver parenchymal diseases could be discriminated by the first 2 variables, and a mean of 86.2% (71-98%) of those with obstructive or space-occupying diseases could be discriminated by the first 3 variables. The basic architecture and function of a quantitative diagnostic system for screening of diseases of the liver and the biliary tract, which we are developing on the basis of this matrix discrimination method, are outlined with our prospects of computer-assisted laboratory diagnosis.