A Systematic Review of Artificial Intelligence-Assisted Chest Radiography for Tuberculosis Case Detection in Primary Health Centres of Maputo Province, Mozambique

Abstract

Tuberculosis (TB) remains a leading cause of mortality in Mozambique, with case detection posing a significant challenge in resource-limited primary health centres. Chest radiography is a key screening tool, but its utility is constrained by a shortage of skilled radiologists. Artificial intelligence (AI)-assisted interpretation of chest X-rays has emerged as a potential solution to augment diagnostic capacity in such settings. This systematic review aimed to evaluate the diagnostic performance, operational feasibility, and implementation challenges of AI-assisted chest radiography for TB case detection within the primary health centres of Maputo Province, Mozambique. A systematic search of multiple electronic databases was conducted for relevant studies. Peer-reviewed articles, reports, and conference proceedings were included. Studies were screened against pre-defined eligibility criteria, with data extracted and synthesised narratively. The methodological quality of included studies was appraised. AI algorithms demonstrated high sensitivity for detecting abnormalities suggestive of pulmonary tuberculosis, with performance often exceeding 90% sensitivity against culture confirmation. Specificity was more variable. Key implementation themes included the critical need for integration into existing clinical workflows, dependency on stable digital infrastructure, and concerns regarding algorithm performance in populations with high HIV co-infection rates. AI-assisted chest X-ray interpretation shows considerable promise for improving tuberculosis triage and screening in Maputo's primary health centres. It could reduce the workload on scarce clinical staff and increase screening throughput, but its effectiveness is contingent on specific contextual implementation factors. Future implementation should prioritise robust pilot studies within routine primary care to validate performance locally. Investment in digital infrastructure and staff training is essential. Development of context-specific algorithms and establishing clear referral pathways for AI-positive cases are also recommended. tuberculosis, artificial intelligence, chest radiography, primary health care, Mozambique, diagnostic accuracy This review synthesises existing evidence on AI for tuberculosis screening in a specific, high-burden sub-Saharan African context, providing pragmatic insights for policymakers and implementers in Mozambique and similar settings.

Ähnliche Arbeiten