Home / Current Issue / Paper 1718207
Artificial Intelligence for Chest Radiograph Analysis in Pediatric Tuberculosis: A Comprehensive Review
Subject area: Biological & Medical Sciences · Area of research: Tuberculosis Diagnosis using AI
DOI: https://doi.org/10.64388/IREV9I11-1718207
Abstract
Tuberculosis (TB) is a leading infectious cause of childhood morbidity and mortality worldwide, with 1.1 million new paediatric cases estimated in 2020. Early detection is paramount, as untreated TB carries a 70% risk of death within ten years. Paediatric diagnosis is uniquely challenging owing to non-specific clinical manifestations, paucibacillary disease, and the difficulty of obtaining adequate respiratory specimens. Chest radiography (CXR) is a highly sensitive, accessible, and affordable screening tool, but accurate interpretation requires considerable expertise that is frequently unavailable in high-burden, resource-limited settings. The World Health Organization (WHO) now recommends the use of computer-aided detection (CAD) software to automate CXR interpretation for TB screening in individuals aged 15 years and older, yet the application of artificial intelligence (AI) to paediatric populations remains nascent. This comprehensive review critically examines the evolution, current evidence, and future potential of AI-driven CXR analysis for paediatric TB. We trace the progression from early conventional CAD systems limited by handcrafted features and modest generalizability to modern deep learning models that automatically learn hierarchical representations from large datasets. Performance metrics are reviewed in depth: while selected algorithms have achieved area under the receiver operating characteristic curve (AUC) values of up to 0.99 in curated datasets, real-world clinical evaluations demonstrate more modest AUCs ranging from 0.71 to 0.94, with sensitivity and specificity varying widely across settings. Only one commercial product, CAD4TB version 6, is currently licensed for use in children over four years. We explore the multifactorial benefits of AI, including enhanced diagnostic accuracy, reduction of interobserver variability, high-throughput screening, augmentation of limited expert capacity, and early detection of subclinical disease. Against these benefits, we highlight persistent challenges: a critical scarcity of large, microbiologically confirmed, and demographically diverse paediatric CXR datasets; the risk of overfitting and dataset-specific bias; variable reference standards; ethical, medicolegal, and regulatory hurdles; and the near-total absence of prospective clinical implementation studies. Future directions including federated learning, lightweight smartphone-deployable models, multimodal diagnostic algorithms integrating clinical and radiological data, and rigorous randomised controlled trials are discussed. With sustained investment, interdisciplinary collaboration, and an emphasis on equity, AI promises to revolutionize paediatric TB diagnosis and contribute meaningfully to the WHO End TB Strategy.
Keywords
Artificial Intelligence, Computer-Aided Detection, Chest Radiography, Paediatric Tuberculosis, Diagnosis, Screening
How to cite this paper
@article{1718207,
author = {Ajiboye Fehinti Prisca, Bukola Ajide},
title = {Artificial Intelligence for Chest Radiograph Analysis in Pediatric Tuberculosis: A Comprehensive Review},
journal = {Iconic Research And Engineering Journals},
year = {2026},
volume = {9},
number = {11},
pages = {4251-4262},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1718207.pdf},
abstract = {Tuberculosis (TB) is a leading infectious cause of childhood morbidity and mortality worldwide, with 1.1 million new paediatric cases estimated in 2020. Early detection is paramount, as untreated TB carries a 70% risk of death within ten years. Paediatric diagnosis is uniquely challenging owing to non-specific clinical manifestations, paucibacillary disease, and the difficulty of obtaining adequate respiratory specimens. Chest radiography (CXR) is a highly sensitive, accessible, and affordable screening tool, but accurate interpretation requires considerable expertise that is frequently unavailable in high-burden, resource-limited settings. The World Health Organization (WHO) now recommends the use of computer-aided detection (CAD) software to automate CXR interpretation for TB screening in individuals aged 15 years and older, yet the application of artificial intelligence (AI) to paediatric populations remains nascent. This comprehensive review critically examines the evolution, current evidence, and future potential of AI-driven CXR analysis for paediatric TB. We trace the progression from early conventional CAD systems limited by handcrafted features and modest generalizability to modern deep learning models that automatically learn hierarchical representations from large datasets. Performance metrics are reviewed in depth: while selected algorithms have achieved area under the receiver operating characteristic curve (AUC) values of up to 0.99 in curated datasets, real-world clinical evaluations demonstrate more modest AUCs ranging from 0.71 to 0.94, with sensitivity and specificity varying widely across settings. Only one commercial product, CAD4TB version 6, is currently licensed for use in children over four years. We explore the multifactorial benefits of AI, including enhanced diagnostic accuracy, reduction of interobserver variability, high-throughput screening, augmentation of limited expert capacity, and early detection of subclinical disease. Against these benefits, we highlight persistent challenges: a critical scarcity of large, microbiologically confirmed, and demographically diverse paediatric CXR datasets; the risk of overfitting and dataset-specific bias; variable reference standards; ethical, medicolegal, and regulatory hurdles; and the near-total absence of prospective clinical implementation studies. Future directions including federated learning, lightweight smartphone-deployable models, multimodal diagnostic algorithms integrating clinical and radiological data, and rigorous randomised controlled trials are discussed. With sustained investment, interdisciplinary collaboration, and an emphasis on equity, AI promises to revolutionize paediatric TB diagnosis and contribute meaningfully to the WHO End TB Strategy.},
keywords = {Artificial Intelligence, Computer-Aided Detection, Chest Radiography, Paediatric Tuberculosis, Diagnosis, Screening},
month = {May},
doi = {https://doi.org/10.64388/IREV9I11-1718207}
}