Artificial Intelligence Applications in Pulmonary Tuberculosis Management

Artificial intelligence (AI) is transforming the landscape of pulmonary tuberculosis (PTB) management by enabling rapid, accurate and resource-efficient approaches to diagnosis, prognosis and treatment monitoring. Integrating machine learning algorithms with radiological, microbiological and clinical datasets has facilitated the development of predictive models that can distinguish Mycobacterium tuberculosis infection from other pulmonary conditions, anticipate treatment response and identify patients at high risk of adverse outcomes. Key AI paradigms include artificial neural networks trained on demographic and symptomatic profiles, support vector machines applied to imaging features and ensemble methods for prognostic scoring. These tools have demonstrated improvements in diagnostic sensitivity over conventional smear microscopy, reductions in time to diagnosis and enhanced stratification of patients for personalised interventions. Globally, AI-driven platforms are being piloted in high-burden settings to support under-resourced health systems, optimise allocation of diagnostic resources and improve adherence monitoring. Ethical deployment and rigorous validation remain priorities to ensure equity and clinical robustness, while continuing advances in explainable AI aim to strengthen clinician trust and regulatory acceptance.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Several recent studies illustrate the breadth of AI applications in PTB care. One foundational investigation deployed artificial neural networks on over 12,000 suspect records, achieving diagnostic accuracy above 94% by integrating demographic, clinical and laboratory variables to predict Mycobacterium tuberculosis positivity. A more resource-constrained setting study explored a machine-learning loop for diagnosis support, training logistic regression, decision trees, random forests, support vector machines and neural networks on seven routinely collected variables; artificial neural networks emerged superior in sensitivity and overall accuracy, outperforming smear microscopy especially in smear-negative cases. In the radiological domain, analysis of computed tomography images annotated by expert radiologists was incorporated into machine-learning pipelines to predict treatment outcomes from initial imaging features. Models balancing class distributions yielded robust performance, and features such as bilateral lung involvement, cavity number and lymph node enlargement were identified as key markers of poor prognosis. Together, these works underscore the utility of AI for rapid triage, risk stratification and therapeutic decision support in diverse healthcare environments.

Artificial Intelligence Applications in Pulmonary Tuberculosis Management publication trend

The graph below shows the total number of articles in artificial intelligence applications in pulmonary tuberculosis management across all publications each year (not limited to Nature Index journals).

Technical terms

Artificial neural network: A computational model composed of interconnected nodes (“neurons”) that learns complex patterns through weighted connections across layers.

Machine learning: A subset of AI in which algorithms improve predictive performance by learning from data rather than following explicit programming rules.

Radiomics: The extraction and analysis of large numbers of quantitative imaging features to characterise disease phenotypes.

Nomogram: A graphical representation of a predictive statistical model that generates a numeric probability of a clinical event.

Area under the receiver operating characteristic curve (AUC): A measure of a model’s ability to discriminate between classes, with 1.0 indicating perfect discrimination and 0.5 reflecting no better than chance.

References

1. Artificial Neural Networks for Prediction of Tuberculosis Disease. Frontiers in Microbiology (2019).
2. Machine learning in the loop for tuberculosis diagnosis support. Frontiers in Public Health (2022).
3. Radiologist observations of computed tomography (CT) images predict treatment outcome in TB Portals, a real-world database of tuberculosis (TB) cases. PLOS ONE (2021).