NOVEL research suggests that computer-aided detection (CAD) software for interpreting chest radiographs (CXR) could play a crucial role in tuberculosis (TB) screening, especially for prevalent TB cases.
The prospective cohort study, conducted in South Africa, followed 483 participants who were household contacts of individuals with rifampicin-resistant TB over a median period of 4.6 years. Participants underwent baseline CXR and sputum investigations, which included both routine and enhanced sputum tests, for prevalent TB.
The study evaluated three CAD software products: CAD4TBv7.0, qXRv3.0.0, and Lunit INSIGHT v3.1.4.111, comparing their performance in detecting TB with blood tests, including the Xpert MTB host-response, erythrocyte sedimentation rate, C-reactive protein, and QuantiFERON.
The study found 23 prevalent TB cases, 7 of which were diagnosed through routine sputum tests, and 38 incident TB cases. The CAD software products showed strong performance in identifying prevalent TB, with the area under the receiver operating characteristic curve (AUC) for CAD4TBv7.0 at 0.87 (95% CI: 0.77–0.96), qXRv3.0.0 at 0.88 (95% CI: 0.79–0.97), and Lunit INSIGHT at 0.91 (95% CIL 0.83–0.99). When compared to blood tests, CAD software showed superior diagnostic performance for prevalent TB.
Interestingly, over 30% of individuals with scores above recommended CAD thresholds who were bacteriologically negative on routine baseline sputum were later diagnosed through enhanced sputum tests or developed incident TB over time. Furthermore, for incident TB cases, the AUC ranged from 0.60–0.65, indicating lower performance.
The authors concluded that, while CAD-CXR shows promise, its utility is not yet fully maximised for diagnosing TB, as many individuals who exceed CAD thresholds but have negative routine sputum tests have true TB disease. Further improvements in CAD thresholds and screening protocols could enhance TB detection and intervention strategies.
Ada Enesco, EMJ
Reference
Macpherson L et al. Diagnostic accuracy of chest x-ray computer-aided detection software for detection of prevalent and incident tuberculosis in household contacts. Clin Infect Dis. 2024; DOI:10.1093/cid/ciae528.
