Implementation of Digital Chest X-Ray with Computer-Aided Detection for Tuberculosis Screening Among Persons with Advanced HIV Disease in Maputo, Mozambique: A Retrospective Cohort Analysis
Authors:
- Maria Ruano CampsMCMaria Ruano Camps
- Bendita JoseBJBendita Jose
- Pereira ZindogaPZPereira Zindoga
- Aleny CoutoACAleny Couto
- Celia CumbeCCCelia Cumbe
- Gil MuvaleGMGil Muvale
- Rosa BeneRBRosa Bene
- Admilson CossaACAdmilson Cossa
- Adrienne E. ShapiroASAdrienne E. Shapiro
- Jeff LaneJLJeff Lane
- Florindo MudenderFMFlorindo Mudender
- Edy NacarapaEdy Nacarapa
Preprint
- 10 June 2026
- preprint
- Published by MDPI AG in Preprints.org
Abstract
Background: Tuberculosis (TB) remains the leading cause of death among persons with advanced HIV disease (AHD) in high HIV-burden settings. Digital chest X-ray (dCXR) with computer-aided detection (CAD) is a promising tool to overcome human resource constraints and improve TB case detection. This study evaluates the real-world im-plementation and performance of dCXR/CAD for TB screening within a specialized AHD clinic in Maputo, Mozambique. Methods: We conducted a retrospective cohort analysis of 487 new AHD patients at Centro de Referência do Alto Maé (CRAM) from October 2023 to September 2024. Of these, 238 underwent dCXR with CAD interpreta-tion. All patients underwent systematic TB screening according to Ministry of Health (MoH) guidelines. Using the recorded diagnosis of TB (bacteriologically confirmed or clinically diagnosed) as the reference standard, we calculated the sensitivity, specifici-ty, positive predictive value (PPV), and negative predictive value (NPV) of the nation-ally adopted CAD threshold (≥0.5). Results: Among 238 AHD patients screened with CAD, 116 (49%) were diagnosed with TB. At the ≥0.5 threshold, sensitivity was 50% (58/116; 95% CI: 41–59), specificity 92% (112/122; 95% CI: 85–96), PPV 85% (58/68; 95% CI: 75–92), and NPV 65.9% (112/170; 95% CI: 58–73). TB diagnosis rates increased sharply with CAD score: 30% (43/143) in normal, 52% (15/27) in abnormal non-suggestive, and 85% (58/68) in suggestive cases. Bacteriological confirmation was low across all groups (19–26%), reflecting reliance on clinical diagnosis. Conclu-sion: Integrating dCXR/CAD into AHD care is feasible and identifies a high TB burden. However, at the adopted threshold of ≥0.5, CAD demonstrated high specificity but low sensitivity (50%) in this population, missing half of all TB cases. These findings suggest that CAD functions better as a confirmatory decision-support tool than a standalone screening test in AHD. Threshold optimization for this specific population warrants prospective evaluation. Implementation challenges — including fragmented systems and lack of dedicated human resources — must be addressed to realize CAD's full po-tential in TB programs.Keywords
All Related Versions
This publication has 22 references indexed in Scilit:
- Incidence and predictors of mortality among persons with rifampicin-resistant tuberculosis and HIV in MozambiqueScientific Reports, 2025
- Evaluating the accuracy of artificial intelligence-powered chest X-ray diagnosis for paediatric pulmonary tuberculosis (EVAL-PAEDTBAID): Study protocol for a multi-centre diagnostic accuracy studyBMJ Open, 2025
- Comparing the accuracy of computer-aided detection (CAD) software and radiologists from multiple countries for tuberculosis detection in chest X-RaysScientific Reports, 2025
- Performance of CAD4TB artificial intelligence technology in TB screening programmes among the adult population in South Africa and LesothoJournal of Clinical Tuberculosis and Other Mycobacterial Diseases, 2025
- Computer-aided detection of tuberculosis from chest radiographs in a tuberculosis prevalence survey in South Africa: external validation and modelled impacts of commercially available artificial intelligence softwareThe Lancet Digital Health, 2024
- Enhanced Tuberculosis Diagnosis With Computer-aided Chest X-ray and Urine Lipoarabinomannan in Adults With HIV Admitted to Hospital (CASTLE Study): A Cluster Randomized TrialClinical Infectious Diseases, 2024
- Extrapulmonary tuberculosis mortality according to clinical and point of care ultrasound features in MozambiqueScientific Reports, 2022
- The Performance of Computer-Aided Detection Digital Chest X-ray Reading Technologies for Triage of Active Tuberculosis Among Persons With a History of Previous TuberculosisClinical Infectious Diseases, 2022
- Predictors of attrition among adults in a rural HIV clinic in southern Mozambique: 18-year retrospective studyScientific Reports, 2021
- Digital Chest Radiography Enhances Screening Efficiency for Pulmonary Tuberculosis in Primary Health Clinics in South AfricaClinical Infectious Diseases, 2021