Application of Pattern Recognition to Knowledge System Design and Diagnostic Inference

An unconventional application of pattern recognition techniques to diagnostic inference is presented. This approach, which is referred to as knowledge-based pattern recognition, is the basis of a diagnostic and consultation system design. The input manifestations are automatically converted to query pattern vectors. The diseases and disease categories are characterized by disease pattern vectors. Diagnostic knowledge is represented in terms of multiple-valued matrices and stored in an associative-tree structure. A variety of similarity measures are considered for pattern matching. Diagnostic inference is performed on the basis of manifestation profile histograms. The proposed approach provides a design technique for knowledge transfer and utilization systems.