"Diagnostic Performance of Artificial Intelligence in Evaluating Tricuspid Regurgitation: A Systematic Review and...

Tricuspid regurgitation (TR) is a common valvular heart disease affecting 0.55%-1.6% of adults, rising to 5%-8% in those over 75 years, often secondary to left-sided pathology or pulmonary hypertension. Moderate-to-severe TR independently predicts mortality and heart failure hospitalization, yet underdiagnosis persists due to echocardiography's operator dependence and variability. Artificial intelligence (AI), including machine learning (ML) and deep learning (DL), promises automated detection to enhance sensitivity and reproducibility. This systematic review and meta-analysis synthesizes evidence on AI's diagnostic performance for TR using echocardiographic or alternative modalities.

Following PRISMA guidelines, we searched PubMed, Embase, Scopus, Web of Science, and EBSCO from inception to August 2025, without language restrictions. Eligibility used PICOS: adults with TR evaluation; AI/ML index tests; clinician-interpreted echocardiography reference; outcomes, including AUROC, sensitivity, and specificity. Data extraction and quality assessment by two reviewers; random-effects meta-analysis for pooled estimates; heterogeneity via I²; certainty per GRADE.

Eight studies were included. Pooled AUROC was 0.89 (95% CI 0.86-0.92) for TR detection. Echocardiography-based models showed sensitivity 0.87 (95% CI 0.81-0.90), specificity 0.88 (95% CI 0.73-0.95), AUROC 0.92 (95% CI 0.89-0.94); ECG-based models had AUROC 0.805. Substantial heterogeneity (I² > 90%) arose from modalities and reference standards; GRADE certainty moderate due to retrospective designs and limited external validation.

AI demonstrates promising diagnostic accuracy for TR, potentially standardizing early detection and triage. However, heterogeneity and methodological gaps necessitate larger prospective, multicenter studies with standardized reporting (e.g., TRIPOD-AI) to confirm clinical utility.