A Vague Expert System Based on Rules for Managing Uncertainty in the Diagnosis of Pulmonary Tuberculosis

Abstract

Pulmonary tuberculosis remains a major public health problem, the diagnosis of which remains difficult due to the variability of clinical manifestations and the inherent uncertainty of medical data, particularly in resource-limited settings. Classical deterministic diagnostic approaches struggle to model this uncertainty, which can limit the reliability and interpretability of computer-assisted clinical decisions. This study proposes the design and evaluation of a fuzzy expert system to facilitate the diagnosis of pulmonary tuberculosis. The system combines a rule-based inference engine with fuzzy logic principles to explicitly represent diagnostic uncertainty and produce graded conclusions. Medical knowledge is formalized as fuzzy rules incorporating linguistic variables such as patient age, symptom duration, and various observed clinical parameters. The system was implemented in the CLIPS environment and evaluated using clinical scenarios inspired by medical practice, demonstrating the feasibility and relevance of the approach. The results show that the fuzzy expert system is capable of providing diagnoses expressed as degrees of certainty, even in situations where traditional expert systems fail to produce a usable decision. The generated treatment recommendations demonstrate satisfactory consistency with established medical protocols. These results confirm that integrating fuzzy logic into an expert system improves the robustness and flexibility of the diagnostic process and constitutes a relevant tool for medical decision support. However, the study is based on simulated scenarios and does not yet rely on real clinical data, which is a limitation and opens up avenues for future validation in a hospital setting, as well as for extending the approach to other pathologies requiring decision-making in uncertain contexts.