Artificial Intelligence in the Diagnosis and Management of Pulmonary Tuberculosis: A Review of Current Applications and Future Perspectives

Abstract

Pulmonary tuberculosis (PTB) remains a major global public health challenge, with traditional diagnostic and management approaches plagued by diagnostic delays, time-consuming drug resistance testing, and subjective efficacy assessment. Artificial intelligence (AI) has emerged as a revolutionary solution for these bottlenecks. This review comprehensively summarizes AI's current applications in PTB care: deep learning models enable automated detection, segmentation and activity differentiation of PTB lesions on chest X-ray/CT with performance comparable to or exceeding human experts (sensitivity >90% for X-ray detection); AI-driven whole-genome sequence analysis rapidly predicts anti-TB drug resistance, shortening testing time from weeks to days; multimodal AI models also show potential in dynamic treatment response monitoring and individualized outcome prediction. However, AI's clinical translation is hindered by data quality/bias, poor model generalizability, low algorithm interpretability, and regulatory/ethical issues. Future priorities include multimodal data fusion, federated learning, prospective clinical validation, and developing lightweight AI models for resource-limited settings. Interdisciplinary collaboration is critical to transform AI from a research tool into a safe, reliable and equitable clinical assistant for PTB care.

Ähnliche Arbeiten