Project Summary
Tuberculosis (TB) remains the leading infectious cause of death globally, with an estimated 10.6 million new cases and 1.3 million deaths reported in 2022. Uganda is among the high TB burden countries, with an incidence of 196 cases per 100,000 population annually. Despite available diagnostic tools, delayed and missed TB diagnoses remain common in low-resource settings due to the limited sensitivity of Ziehl-Neelsen (ZN) smear microscopy, dependence on skilled personnel, and restricted access to molecular testing such as GeneXpert.
This study aims to validate the diagnostic accuracy of an artificial intelligence (AI)–driven sputum analysis platform for laboratory-based TB detection in South-Western Uganda. Using digitized archived ZN-stained sputum smears from Mbarara Regional Referral Hospital, the platform will employ deep learning architectures (ResNet and DenseNet) to automatically detect Mycobacterium tuberculosis bacilli. The diagnostic performance of the AI system will be evaluated against conventional ZN microscopy by estimating sensitivity, specificity, and predictive values.
A minimum of 512 archived sputum smears will be analyzed using a laboratory-based analytical validation design. Discrepant results will undergo blinded expert review. This study will generate locally relevant evidence on the accuracy, reliability, and feasibility of AI-assisted sputum microscopy under real-world laboratory conditions. Findings are expected to inform the integration of AI-powered diagnostics into Uganda’s TB control program, supporting faster, standardized, and scalable TB detection in resource-limited settings and aligning with WHO End TB targets.
Lead Principal Investigator: Habert Tumwesigye
Co- Principal Investigator: Dr Atwine Daniel.
Implementers: Collaborative project between Mbarara University of Science and Technology and SRF Research and Training Centres, Mbarara, Uganda.
Funding Source: MUST Internal Research funds from Government of Uganda-DRGT
Current status: Ongoing regulatory approvals
Duration: 6 months
Start date: October 2025
Expected End date: May-June 2026