THAA0305 | AI-DRIVEN SCREENING IN TB/HIV PROGRAMMING: ADVANCING EARLY DETECTION WITH CAD TB IN NIGERIA’S ACTIVE CASE FINDING

Paul Alu, O. Daniel, Pedro Michael, Aderonke Agbaje, Charles Mensa, Patrick Dakum | Ihvn, Lagos, Nigeria

The use of artificial intelligence (AI) in TB screening and diagnosis, as recommended by the WHO, has revolutionized the detection of TB, particularly in high-burden countries like Nigeria. Computer-aided detection (CAD) using AI to analyze portable digitally captured chest X-ray (PDX) images improves the sensitivity and speed of detecting abnormalities suggestive of pulmonary TB. This technology has become especially critical in identifying TB among people living with HIV (PLHIV), whose TB presentation may be atypical and easily missed by conventional methods. Given that TB remains the leading cause of death among PLHIV, integrating AI-enabled CAD4TB into TB/HIV collaborative activities can significantly reduce missed diagnoses. This study assessed the contribution of AI-powered CAD4TB in TB case finding during active case finding (ACF) interventions in Nigeria, while also examining its added value in improving TB screening among HIV-positive individuals. We conducted a retrospective analysis of 36 months of data (January 2022 – December 2024) from the USAID TB-LON 3 project in Oyo State, Nigeria. Implemented by the Institute of Human Virology Nigeria (IHVN), the study compared AI-enabled CAD4TB PDX machine performance with conventional symptom-based screening. Data included the general population and PLHIV screened in facility and community settings. A total of 66,861 individuals were screened using the AI-PDX machine, identifying 6,117 presumptive TB cases, with 1,196 confirmed diagnoses. Among the confirmed cases, 14% were PLHIV, demonstrating higher detection efficiency in co-infected individuals. In comparison, conventional screening evaluated 707,046 people, identifying 70,680 presumptive cases and 5,158 TB diagnoses. The AI-PDX method yielded a lower number needed to test (NNT: 55) and number needed to screen (NNS: 5) versus conventional screening (NNT: 136; NNS: 14). TB case finding increased by 24% post-CAD4TB implementation. The integration of AI-enabled CAD4TB PDX machines significantly enhances TB case finding efficiency, especially among populations with high TB/HIV co-infection rates. This approach is not only cost-effective but also crucial in reducing missed TB cases among PLHIV, who often present diagnostic challenges. Given the dual burden of TB and HIV in Nigeria, scaling up this innovation within both TB and HIV programs is essential for bridging diagnostic gaps.