Artificial intelligence applications relevant to infective endocarditis in an oral–systemic context: a scoping review

Abstract

Infective endocarditis (IE) remains associated with substantial morbidity and mortality, and its early diagnosis and risk stratification can be challenging. The oral cavity may contribute to bacteremia through oral microorganisms, including viridans group streptococci, during routine daily activities and dental procedures. Artificial intelligence (AI) is increasingly being investigated for diagnostic and predictive applications in IE; however, the extent to which oral and dental data are represented in these models has not been clearly mapped. This scoping review was conducted in accordance with PRISMA-ScR guidance and structured using the Population–Concept–Context (PCC) framework. PubMed, Scopus, and the Cochrane Library were searched for English-language studies published between 2020 and 2025. Eligible studies applied AI or machine learning methods to infective endocarditis, bacteremia or bloodstream infection, or oral and dental infections with potential relevance to oral–systemic infectious pathways. Ten studies met the inclusion criteria. Most used retrospective designs and clinical datasets and evaluated machine learning approaches such as Random Forest and XGBoost. Reported AUC values were generally above 0.80, although study designs, outcomes, and validation approaches varied considerably. Oral health–specific data were rarely incorporated into model development. Only one study directly integrated structured dental data, including panoramic radiographs and clinical oral variables, whereas most studies relied primarily on conventional medical data. Internal validation was more common than external validation. AI has been applied to several aspects of IE diagnosis, risk prediction, and outcome assessment, but the integration of oral and dental data remains limited. Future studies should evaluate models that combine dental, microbiological, imaging, and systemic clinical information and should place greater emphasis on external validation and clinical applicability.