Advanced tuberculosis diagnosis system: Integrating case-based reasoning with nearest neighbor algorithm.

BACKGROUND

A serious infectious illness with a high morbidity and death rate worldwide, tuberculosis (TB) is more prevalent in low- and middle-income nations. Although there are a number of diagnostic techniques, the most only address tuberculosis in the lung and ignore drug-resistant strains (MDR-TB, XDR-TB) as well as tuberculosis lymphadenitis. A thorough diagnostic system that covers all types of tuberculosis is essential.

OBJECTIVES

To enhance TB diagnosis, particularly pulmonary TB, lymphadenitis, and drug-resistant TB, this study offers an expert system based on Case-Based Reasoning (CBR) and the Nearest Neighbor Algorithm.

METHODS

Information was gathered from hospital records of prior tuberculosis cases, including 43 cases from Debre Tabor General Hospital. In addition to document analysis, information was acquired through both structured and unstructured interviews with medical specialists. The R4 model-Retrieve, Reuse, Revise, and Retain-is followed by the system architecture. Recall, expert acceptance, and precision were among the evaluation metrics.

RESULTS

The system had an 86.5% expert acceptance rate, 84.7% precision, and 75.3% recall. Compared to previous medical diagnostic methods, it shown a notable improvement, especially in diagnosing mental health and hypertension.

CONCLUSION

By combining Case-Based Reasoning and the Nearest Neighbor Algorithm, it is possible to diagnose tuberculosis (TB) more effectively and with greater accuracy. This integration also makes it possible to diagnose cases that are resistant to drugs. In order to improve the system's performance even more, future research may investigate the integration of additional reasoning strategies.