Abstract
BACKGROUND: India accounts for the highest Tuberculosis (TB) burden globally. The incidence and prevalence of TB are higher in tribal population than general population. In this study, we assessed the effectiveness of artificial intelligence (AI) based chest X-ray (CXR) interpretation software device (qXR version 3), in detecting TB from a predominantly tribal population setting. METHODS: In this multicenter prospective study, all the CXRs of patients aged > 15 years taken for any reason at 3 public health facilities in the Chhattisgarh state of India between 01 August 2023 and 31 March 2024 were included. Patients flagged by AI as TB presumptive were directed to undergo sputum testing, who are subsequently confirmed either microbiologically or clinically. RESULTS: Out of 2745 CXRs screened, 363 patients (median age, 44 years [IQR: 30-53]; 261 [71.9%] male) were identified as presumptive for TB. 162 cases were confirmed with TB positivity rate of 44.63% (95% CI: 39.44-49.91). Among the AI-flagged cases, 51 (14.04%) patients were asymptomatic, and 20 (39.22%) of them were confirmed with TB. Descriptively, when compared with baseline (August-2022 to March-2023), an 80.21% (P < .001) increase in the number of TB case notifications was observed during the AI implemented period. CONCLUSIONS: This study highlights the potential of AI to enhance TB detection and feasibility in a resource-limited tribal setting. Above 40% of the patients flagged by AI were subsequently confirmed to have the TB disease. Additionally, the study demonstrated the potential of AI in identifying asymptomatic individuals who would otherwise have been missed or diagnosed late.