Automated assessment of left ventricular ejection fraction in the emergency department using a pocket ultrasound device: a reproducibility study

Abstract Background Left ventricular ejection fraction (LVEF) is a key parameter in focused cardiac ultrasound for the bedside assessment of emergency department (ED) patients. Pocket ultrasound devices (PUDs) now offer automated LVEF estimation via embedded algorithms, but the reproducibility of such methods in emergency settings remains poorly established. Methods This prospective single-center study assessed the inter- and intra-operator reproducibility of automated LVEF measurements using a commercially available PUD (Vscan Extend™) in a tertiary care ED. Two trained emergency physicians (EPs) independently acquired two apical four-chamber cine loops per patient. If the acquisitions were successful, automated and visual LVEF were estimated from each loop. The primary outcome was inter-operator reproducibility of automated LVEF estimation, assessed by Lin’s concordance coefficient. Secondary outcomes included intra-operator reproducibility, visual LVEF reproducibility, feasibility, and the influence of image quality. Results Among 119 patients, inter-operator reproducibility of automated LVEF was poor (Lin’s coefficient 0.23; 95% CI: 0.01–0.44), and intra-operator reproducibility was similarly limited (0.46; 95% CI: 0.33–0.58). Visual LVEF estimation showed higher agreement (inter-operator: 0.87; intra-operator: 0.98). Visual LVEF estimation was feasible in 117 cases (98%; 95% CI: 96–100), while automated estimation was feasible in 89 cases (75%; 95% CI: 67–83), p < 0.01. Automated LVEF estimation was feasible in 63 of 74 cases with good or excellent image quality (85%) and in 26 of 45 cases rated as having insufficient or barely sufficient image quality (58%) ( p < 0.01). Conclusion In this emergency department setting, the automated LVEF workflow evaluated in this study showed poor reproducibility, supporting visual estimation by trained emergency physicians as the preferred method for bedside LVEF assessment. Future research should focus on improving automated image acquisition and analysis to reduce measurement variability before routine clinical implementation. Trial registration NCT04366453, 2020-11-18.