Automated Echocardiographic Quantification of Left Ventricular Ejection Fraction Without Volume Measurements Using a...

Area of Science:

• Cardiology
• Artificial Intelligence in Medicine
• Medical Imaging Analysis

Background:

• Echocardiography relies on manual or automated endocardial boundary identification for left ventricular ejection fraction (LVEF) calculation.
• Current automated methods can be prone to errors, limiting accuracy in some patients.
• A novel approach is needed to circumvent border detection for more reliable LVEF estimation.

Purpose of the Study:

• To develop and validate a fully automated machine learning algorithm for LVEF estimation.
• To assess if the algorithm can estimate ventricular contraction directly, bypassing traditional border detection.
• To compare the accuracy of the automated method against expert measurements.

Main Methods:

• A machine learning algorithm (AutoEF) was trained on over 50,000 echocardiographic studies.
• The algorithm was tested on 99 independent patients, comparing automated LVEF to expert-averaged measurements.
• Statistical analyses included linear regression, Bland-Altman analysis, and assessment of sensitivity and specificity for detecting low LVEF.

Main Results:

• Automated LVEF estimation was feasible in all patients with high consistency (mean absolute deviation 2.9%).
• AutoEF demonstrated excellent agreement with reference values (r=0.95) and high sensitivity (0.90) and specificity (0.92) for detecting LVEF ≤35%.
• Performance was comparable to that of clinical readers.

Conclusions:

• A volume-independent machine learning algorithm for LVEF estimation is highly feasible.
• The automated method achieves accuracy comparable to conventional volume-based measurements.
• This AI-driven approach offers a promising alternative for accurate LVEF assessment.