Pattern Recognition Approach to Medical Diagnosis

Abstract

A sequential method of pattern recognition was used to recognize hyperthyroidism in a sample of 2208 patients being treated at the Straub Clinic in Honolulu, Hawaii. For this, the method of class featuring information compression (CLAFIC) [1] was used, introducing some significant improvements in computer medical diagnosis, which by its very nature is a pattern recognition problem. A unique subspace characterizes each class at every decision stage, and the most prominent class features are selected. Thus the symptoms which best distinguish hyperthyroidism are extracted at every step and the number of tests required to reach a diagnosis is reduced.