[Optimization of the decision-making guidelines for the diagnosis of cerebrovascular diseases].

The problem to optimize decision rules and the number of informative signs for the automated diagnosis of the cerebrovascular pathology is analysed. A double application of the algorithm to subdivide the set into clusters makes it possible to decrease the number of informative counts and to rearrange separate signs thus decreasing the dimensionality of the sum dividing subspace and improving a recognition quality for a newly designed classifier even though the number of classes being diagnosed increases.