AI System Enables the Detection of TB in Children

3 November 2025

Researchers from the Universidad Politécnica de Madrid (UPM) and the Centro de Investigación Biomédica en Redes, Bioingeniería, Biomateriales y Nanomedicina (CIBER-BBN), in collaboration with international institutions and organizations, developed an artificial intelligence (AI) system which helps to detect radiological signs consistent with pulmonary tuberculosis in pediatric chest X-rays. The work, published in Nature Communications, is the first study which systematically evaluate the value of lateral chest X-rays in this context and compare age-specific models against general models trained on all ages.

Among the collaborators stand out the Instituto de Salud Global de Barcelona (ISGlobal), a center supported by the Caixa Foundation, the Centro de Investigação em Saúde de Manhiça (CISM) in Mozambique, the Red Española de Estudio de Tuberculosis Pediátrica (pTBred), the Centro de Investigación Biomédica en Red de Enfermedades Infecciosas (CIBERINFEC) and the Children´s National Hospital (Washington DC, USA).

Tuberculosis in children represents a diagnostic challenge, with often nonspecific and radiological variations tend to be more subtle and variable compared to adults. To deal with these difficulties, the system integrates frontal chest X-rays and, when available, lateral X-rays as well. This system has been optimized for efficiency, trained and validated using data from various hospitals and epidemiological contexts.

Daniel Capellán, first author and UPM researcher states: “we have designed this tool to be extremely efficient without sacrificing accuracy or performance, with the aim of being integrated even into mobile devices and thus bringing tuberculosis diagnosis to rural areas with a high incidence of the disease, where resources and access to specialized radiologists are very limited. “Juan José Gómez, second author of the study and UPM professor adds: “Pre-training on adult data allows us to take advantage of more diverse datasets, which makes it easier for the model to learn robust features that can then be adapted to the pediatric context”.

Main contributions

This work makes three key contributions. First of all, it demonstrates that pre-training AI models on large collections of adult radiographs improves performance when these models are further refined using pediatric data.

Secondly, it highlights the utility of lateral radiographs, which offers particularly valuable complementary information in infants and young children, where the frontal view may be insufficient. Third, it shows that age-specific models outperform models trained on all ages, due to differences in development and clinical disease presentation across different age groups.

Elisa López, a researcher at ISGlobal during the study, emphasizes: “Lateral views complement the frontal view and are specially valuable in infants and young children, as they help to identify discoveries that might be missed when only one projection is available”.

Implications for clinical practice and public health

“This solution does not try to replace radiologists or physicians, but rather to serve as a support tool: it can help to prioritize studies, guide screening decisions as well as facilitate early detection in resource-limited settings”, says Begoña Santiago, coordinator of pTBred and pediatrician at Hospital General Universitario Gregorio Marañón de Madrid. “The use of lateral views and their adaptation by age group could increase diagnostic awareness in pediatric populations, especially in infants and young children, in which diagnosis is more complex”.

Acknowledgments

This research was possible thanks to the support of a broad network of pediatricians in Spain and Mozambique, as well as the support of the Ministerio de Ciencia e Innovación, the Instituto de Salud Carlos III, the Sociedad Española de Neumología y Cirugía Torácica (SEPAR) and several European projects funded by the European Union, such as INNOVA4TB, ADVANCETB and Stool4TB. The authors also gratefully acknowledge the funding and collaboration of the European Respiratory Society and the contributions of all participating centers and researchers, including ISGlobal, CISM, pTBred, CIBERINFEC and Children´s National Hospital.

Capellán-Martín, D., Juan J. Gómez-Valverde, …, Maria J. Ledesma-Carbayo et al. (2025). Multi-view deep learning framework for the detection of chest X-rays compatible with pediatric pulmonary tuberculosis, Nature Communications. https://doi.org/10.1038/s41467-025-64391-1
Gómez-Valverde, J. J., Sánchez-Jacob, R., Ribó, J. L., Schaaf, H. S., García Delgado, L., Hernanz-Lobo, A., … & Ledesma-Carbayo, M. J. (2024). Chest X-Ray–Based Telemedicine Platform for Pediatric Tuberculosis Diagnosis in Low-Resource Settings: Development and Validation Study. JMIR pediatrics and parenting, 7, e51743. https://doi.org/10.2196/51743
Capellán-Martín, D., Gómez-Valverde, J. J., Bermejo-Peláez, D., & Ledesma-Carbayo, M. J. (2023, April). A lightweight, rapid and efficient deep convolutional network for chest X-ray tuberculosis detection. In 2023 IEEE 20th International Symposium on Biomedical Imaging (ISBI)(pp. 1-5). https://doi.org/10.1109/ISBI53787.2023.10230500

Source: Universidad Politécnica de Madrid