A Prototype Diagnostic Expert System for Common Respiratory Diseases Using Dempster Shafer Theory

Abstract

Medical diagnostic investigations arc inherently very compkx. The doctm is faced \\ith a patient who has his own personal experiences, knowledge from books. and beliefs. The doctor notes tl1l' patient" s signs and ~.:mptoms, combines these with thL· patient's medical history. physical examination and laboratory lindings and then diagnoses the disc1sc. Medical decision-making and particularly the establishment of a diagnosis is an l'tTllrprone process. Computers can be used etTectively to assist the physicians in the process of diagnosis. Various methods have been followed in order to computerize the process o 1· medical diagnosis fl ]. !\prototype medical diagnostic system was developed to diagnose five rl·spiratory diseases using fuzzy logic in [2]. Fuzzy logic had been identified to have a k\v drawbacks when used for medical diagnosis as suggested in [3 ]. Dempster Shafer theory provides a suitable framework for the incorporation of medical knmvledge. As suggested in [3] Dempster Shafer theory provides a method of using evidential reasoning for diagnostic inference. In this paper a decision support mechanism is attempted utiltzing tive respiratory related diseases \Vith similar presenting clinical features. w·hich arc hard to tell apart tor the non expert. The medical knowledge in the system is represented by Basic Probability Assignment (BPA) values pertaining to all the clinical features and diseases considered. The inference mechanism relies on the Dempster's rule of combination. The output of the system is diagnostic hypotheses along with a belief measure for each disease. The belief values are calculated using limiting functions based on belief and plausibility. A clinical validation ofthe system is currently underway at the ·y caching Hospital Peradeniya (THP). The preliminary results have proved to be conducive and identified shortcomings are to be addressed in the future.