Conference Abstract | Volume 8, Abstract NACNDC/19JASH077 (Oral) | Published: 26 Nov 2025
Clark Joshua Brianwong1, Cherotin Diana1, Lwanga Sssekiswa Zimwanguyiza1, Bakyawa Jennifer1, Richard Jjuuko1, Patricia Nahirya1, Alexander Mugume1, Dithan Kiragga1
1Baylor Foundation Uganda
&Corresponding author: Clark Joshua Brianwong, Baylor Foundation Uganda, Email: clarkjoshua001@gamil.com, ORCID: https://orcid.org/0009-0005-1548-2684
Received: 30 Sep 2025, Accepted: 20 Oct 2025, Published: 26 Nov 2025
Domain: Infectious Disease Epidemiology
Keywords: Data driven targeting; AI-assisted, digital Xray; tuberculosis; eCBSS; mobile screening
©Clark Joshua Brianwong et al. Journal of Interventional Epidemiology and Public Health (ISSN: 2664-2824). This is an Open Access article distributed under the terms of the Creative Commons Attribution International 4.0 License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Cite this article: Clark Joshua Brianwong et al., Data-driven deployment of an AI mobile CXR van to accelerate TB detection in Eastern Uganda. Journal of Interventional Epidemiology and Public Health. 2025;8(ConfProc6):077.https://doi.org/10.37432/JIEPH-CONFPRO6-0077
Introduction: Eastern Uganda continues to miss a sizeable share of tuberculosis (TB) cases (about 40%) despite expanded facility-based screening. Routine surveillance through the electronic Case-Based Surveillance System (eCBSS) and digital contact tracing provides granular spatiotemporal signals but remains underused for directing high-yield interventions. We used a data-driven approach that fused eCBSS and other routine “big data” streams to guide dynamic deployment of an AI-assisted mobile chest X-ray (CXR) van to accelerate TB detection.
Methods: During the September 2024 CAST+ campaign, we integrated eCBSS event data (case notifications and GPS), digital contact-tracing outputs, EMR extracts and community active-case-finding logs into a harmonized data mart. A hotspot engine (kernel-density surfaces with time-weighted clustering) generated weekly micro-catchment risk maps at the parish/village level. An optimization layer ranked outreach sites by expected yield, combining historical positivity and access. The mobile van conducted community CXR with on-site CAD/AI triage; clients exceeding predefined AI thresholds provided sputum for GeneXpert per NTLP guidance. A digitalized workflow supported consent, registration, geotagging, referrals, and linkage.
Results: Across 92 mapped hotspots, 2,424 community CXR screenings were completed, with 417 abnormal reads and 178 TB cases diagnosed and started on treatment. Yield in X-ray–defined hotspots was 11%, markedly higher than symptom-only screening streams (1.2%; 69 cases among 5,503 presumptives of the 38992 screened), despite similar community contexts. The integrated workflow supported rapid referral and traceability across registration, testing, and treatment initiation.
Conclusion: Integrating routine surveillance with an AI-enabled mobile CXR platform and geospatial targeting increased TB detection in Eastern Uganda. This precision model directs resources to high-need micro-catchments, strengthens real-time oversight, and scales within NTLP systems. Broader adoption could raise notifications, shorten time-to-treatment, and accelerate End TB progress where access barriers and delays persist.
Menu