Diagnostic Potential of AI-driven mRNA Host-Response in Bacterial LRTI

Résumé fourni par la source

Lower respiratory tract infections (LRTI) are a major global health burden, yet its diagnosis remains challenging as no reliable reference standard exists. Artificial intelligence (AI) has the potential to enhance diagnostic accuracy. This study explores the association between clinical presentation, microbiology, and AI-based 29-mRNA host-response in patients with suspected LRTI. Methods: Prospective, single-center observational study, 578 patients undergoing bronchoscopy with BAL. A 29-mRNA host-response test (InSep™, Inflammatix) with AI-driven interpretation (IMX-BVN-3) was performed on blood samples. Adjudication of cases by clinical, laboratory, radiological, and microbiological data by specialist panel. BAL sample analyis by conventional and molecular microbiological methods. Comparison of AI-based mRNA to clinical and microbiological results. Results: Mean patient age 64±1 years, 46.7% immunocompromised. Clinical presentation significantly differentiated bacterial infection probability based on mRNA signatures (p<0.0001), microbiological methods did not (p=0.1987). Using conventional methods as a reference, AI-based mRNA host-response score was 32.7% sensitivity, 75.9% specificity for bacterial infection, compared to clinical adjudication with 67.2% sensitivity, 56.2% specificity. Conclusions: mRNA host-response testing offers a novel approach diagnosing bacterial LRTI. Combination of clinical assessment, microbiology, and AI-based host-response testing may improve diagnostic accuracy. Factors e. g. immunosuppression and bacterial colonization must be considered when interpreting results. Further research is needed to establish AI-assisted tests to improve diagnostic acccuracy of bacterial LRTI.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.