A Belief Rule Based Expert System to Assess Tuberculosis under Uncertainty

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Expert Systems

Background:

• Tuberculosis (TB) diagnosis relies on signs and symptoms, which are subject to significant uncertainty (vagueness, imprecision, randomness).
• Traditional diagnostic methods by physicians can yield unreliable results due to inherent uncertainties in symptom interpretation.
• Existing fuzzy rule-based systems may not fully address the multifaceted nature of diagnostic uncertainty in TB.

Purpose of the Study:

• To design, develop, and apply a Belief Rule Based Expert System (BRBES) for Tuberculosis (TB) diagnosis.
• To create a system capable of effectively handling various types of uncertainties present in TB diagnostic data.
• To enhance the reliability and accuracy of primary TB diagnosis through an AI-driven approach.

Main Methods:

• Constructed a knowledge base for the BRBES using expert opinions and historical TB patient data.
• Developed a Belief Rule Based Expert System (BRBES) incorporating mechanisms to manage uncertainty.
• Conducted experiments using data from 100 TB patients to evaluate system performance.

Main Results:

• The BRBES achieved more reliable diagnostic outcomes than human expert assessments.
• The BRBES outperformed a fuzzy rule-based expert system in TB diagnosis accuracy.
• Experimental results indicate the BRBES's efficacy in handling diagnostic uncertainties.

Conclusions:

• The developed BRBES offers a more reliable approach to Tuberculosis diagnosis compared to traditional methods.
• The system's ability to manage uncertainty makes it a valuable tool for improving diagnostic accuracy.
• BRBES represents a significant advancement in the application of AI for medical diagnosis, specifically for TB.