Left ventricular ejection fraction assessment: artificial intelligence compared with echocardiography expert and cardiac magnetic resonance measurements.

Abstract

Introduction:  Cardiac magnetic resonance (CMR) is the gold standard for assessing left ventricular ejection fraction (LVEF). Artificial intelligence (AI)-based echocardiographic analysis is increasingly utilized in clinical practice.

Objectives:  This study aimed to compare the results of LVEF echocardiography (ECHO) assessed by experts and automated AI, with CMR as the reference standard.

Patients and methods:  We retrospectively analyzed 118 patients who underwent both CMR and ECHO within 7 days. LVEF measured on CMR was compared with the results obtained from an AI‑ based software, which automatically analyzed all stored digital imaging and communications in medicine loops (multiloop AI analysis) on ECHO. Additionally, AI results were repeated using only 1 best‑ quality loop for 2-chamber view and 1 for 4‑ chamber view (single loop AI analysis) on ECHO. These results were further compared with standard ECHO analysis performed by 2 independent experts. Agreement was investigated using the Pearson correlation and Bland-Altman analysis as well as the Cohen κ and concordance for categorization of LVEF into subgroups (≤30%, 31%-40%, 41%-50%, 51%-70%, and >70%).

Results:  Both experts demonstrated strong inter‑ reader agreement (R = 0.88; κ = 0.77), and their assessment correlated well with CMR‑ assessed LVEF (expert 1, R = 0.86; κ = 0.74; expert 2, R = 0.85; κ = 0.68). The results of the multiloop AI analysis correlated strongly with those of CMR (R = 0.87; κ = 0.68) and the experts (R = 0.88-0.9; κ = 0.77). The single‑ loop AI analysis demonstrated numerically higher concordance with CMR‑ assessed LVEF (R = 0.89; κ = 0.75) than the multiloop AI analysis and expert analysis.

Conclusions:  AI‑ based analysis showed similar LVEF assessment results as human expert analysis in comparison with CMR. AI‑ based ECHO analysis is a promising approach, but its results should be interpreted with caution.