Combining medical records with case-based reasoning in a mixed paradigm design — TROPIX architecture & implementation

Abstract

The design architecture, case representation of actual medical cases, and the implementation of TROPIX, a tropical disease diagnosis and therapy tool are presented. TROPIX is intended to be bundled with laptop computers for rural health centers and semi-urban hospitals in developing countries. Our model uses the ideal domain knowledge to build a decision matrix, DM from disease features against which new cases are measured using the MVF (Matched Vector Functions) algorithms. The result of the diagnosis stage provided by the MVF procedure and other parameters of the features are then used for CBR case retrievals, verification, and case validation. The final stage in the design yields the selection of 1 or 2 competing cases presented for reuse and perhaps for subsequent repair and adaptation.

The solution for the new case solved is either one or more actions, a therapy plan or recommendations. The design demonstrates how we can integrate domain knowledge and (medical) records with CBR, statistical pattern recognition, and abductive reasoning to build a knowledge based system.

Part of the Ph.D. research effort of the author at the FUTO with benchwork done at the CACS (1993–97).

I remain indebted to Eng. Prof. C.O.G. Obah, the Vice-Chancellor of FUTO, also my mentor, who charted a course for me in academic endeavors.