Project Details

Performance of computer aided detection for tuberculosis diagnosis in children living with HIV

Anca Vasiliu, MD, MPH, PhD

Summary

BACKGROUND: Pediatric tuberculosis poses a significant global health challenge, with 63% of TB cases below 15 years of age not diagnosed or unable to access TB diagnostic and treatment services. Chest radiography (CXR) remains extensively utilized in clinical decision-making for pediatric tuberculosis, especially when microbiological tests yield negative results or are inaccessible and artificial intelligence can read CXRs using computer aided detection (CAD) that provides a timely and accurate score indicating the probability of pulmonary TB. GAP: CAD TB screening algorithms have not been validated in children and CAD algorithms have not been developed to identify the unique CXR manifestations seen in CLHIV who have TB. HYPOTHESIS: We hypothesize that an optimized CAD reading score specific for CLHIV will increase the TB diagnostic yield and that the AI software is concordant with human readers. We hypothesize that the TB treatment outcomes of CLHIV diagnosed using CAD technology are better than the TB treatment outcomes of CLHIV in the year prior to the project. METHODS: This is a cohort study of diagnostic accuracy conducted among CLHIV evaluated for TB at PMGH. All CLHIV <15 years of age consulting in the HIV clinic and completing evaluation for TB at PMGH will be included in the study. RESULTS: Pending. IMPACT: CLHIV worldwide would benefit from an accurate pulmonary TB CAD score specifically calibrated for them. Implementation of CAD-based CXR diagnostics in CLHIV evaluated for TB at PMGH could lead to earlier diagnosis, improved access to care, optimal clinical outcomes, cost savings, and enhanced healthcare delivery.