Multi-view deep learning framework for the detection of chest X-rays compatible with pediatric pulmonary tuberculosis.

Tuberculosis (TB) remains a major global health burden, particularly in low-resource, high-prevalence regions. Pediatric TB diagnosis poses challenges with non-specific symptoms and less distinct radiological manifestations than adult TB. Many affected children remain undiagnosed or untreated. The World Health Organization (WHO) recommends chest X-ray (CXR) for TB screening and triage, given its accessibility and rapid assessment of pulmonary TB-related abnormalities. We present pTBLightNet, a multi-view deep learning framework to detect pediatric pulmonary TB by identifying TB-compatible CXRs with consistent radiological findings. Leveraging both frontal and lateral CXR views, our framework is pre-trained on adult CXR datasets (N = 114,173), then fine-tuned or trained from scratch, and subsequently evaluated on CXR datasets (N = 918) from three pediatric TB cohorts. It achieves an area under the curve (AUC) of 0.903 and 0.682 on internal and external testing, respectively. External evaluation supports its effectiveness and generalizability using CXR TB compatibility, expert reading, microbiological confirmation and case definition as reference standards. Age-specific models (<5 and 5-18 years) perform competitively with those trained on larger undifferentiated populations, and adding lateral CXRs improves diagnosis in younger children. These results highlight the robustness of our approach across age groups and its potential to improve TB diagnosis, particularly in resource-limited settings.