Abstract 4372979: AI-Powered Handheld Ultrasound for Rapid Left Ventricular Assessment: A Systematic Review and Meta-Analysis

Background: Handheld ultrasound (HHU) is increasingly used in emergency and point-of-care settings for cardiac assessment. The integration of artificial intelligence (AI) into HHU offers the potential to improve diagnostic accuracy and speed in evaluating left ventricular (LV) systolic function. We conducted a systematic review and meta-analysis to determine the diagnostic performance of AI-enabled HHU in identifying reduced ejection fraction (EF). Methods: PubMed, Scopus, and Embase were searched through April 2024 using terms: (“AI” OR “machine learning”) AND (“handheld ultrasound” OR “point-of-care ultrasound”) AND (“left ventricular function” OR “ejection fraction”). Inclusion criteria: (1) studies comparing AI-enhanced HHU with echocardiography reference standard; (2) reported AUC, sensitivity, or specificity for detecting EF <40%; (3) conducted in humans. Data were extracted and pooled using a random-effects model. Subgroup analysis examined performance by device type and clinician experience. Heterogeneity was assessed using I2, and study quality by QUADAS-2. Results: Nine studies (n = 4,117 patients) met criteria. The pooled AUC for AI-HHU detection of EF <40% was 0.92 (95% CI: 0.89–0.94), with sensitivity 88% and specificity 90%. Subgroup analysis showed higher accuracy in AI-assisted novice users (AUC = 0.94) vs experts (AUC = 0.89). AI models based on convolutional neural networks (CNNs) consistently outperformed traditional regression models. In a pooled comparison across three studies, AI models had significantly lower interobserver variability (mean difference = 0.07, p<0.01). No publication bias was detected (Egger’s p = 0.27). Conclusion: AI-enhanced HHU provides accurate, reproducible LV function assessment, particularly for detecting reduced EF. Its use may enhance clinical workflows, especially among non-experts in emergency care settings. These findings support broader integration of AI tools into bedside imaging platforms for rapid triage and management of cardiac patients.