Knowledge representation and compilation for symptom-disease-testrelationships

A method that represents and compiles expert knowledge for choosing diagnostic tests is presented. Usually, physicians inquire into the patient's complaints, examine him/her physically, and on the assumption of a specific disease or impaired organ system, choose a certain group of laboratory tests designed to examine this assumption. After checking the test results, they may perform another set of tests for further differential diagnosis. This step is repeated until they obtain the final diagnosis. In the proposed method, this knowledge is categorized into three relations. The relation SD is defined on domains symptom/sign and disease, the relation DT is on disease and diagnostic test, and the relation DD is on disease and disease where the first attribute (column) in each relation, respectively, is a primary key, that is, there exists precisely one value, and the second is a set of values. We show that redundant tuples can be reduced or decomposed by checking whether or not there exist tuples whose set of diagnostic tests are equivalent to that of other tuples or are a subset of others. Finally, we can obtain a preferable set of diagnostic tests for given symptoms/signs by using the join of SD and DT, and for further diagnosis, the join of DD and DT is applicable.