Assessment and validation of a novel fast fully automated artificial intelligence left ventricular ejection fraction...

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Left ventricular ejection fraction (LVEF) quantification via transthoracic echocardiography (TTE) is often subjective, time-intensive, and prone to errors.
• LVivoEF is an AI-powered software designed for rapid LVEF assessment by automatically tracking endocardial borders.

Purpose of the Study:

• To evaluate the accuracy of LVivoEF in quantifying LVEF.
• To compare LVivoEF's performance against physician-measured LVEF (MD-EF) and cardiac magnetic resonance imaging (cMRI) as the reference standard.

Main Methods:

• A comparative study involving 273 patients, where AI-derived LVEF and MD-EF were assessed against cMRI.
• AI-derived LVEF was generated from a standard four-chamber view without manual adjustments or ultrasound enhancing agents (UEAs).
• Exclusion criteria included interval interventions, incomplete studies, or uninterpretable views, resulting in a final analysis of 242 subjects.

Main Results:

• AI-derived LVEF demonstrated strong correlation with cMRI (r=0.890), comparable to MD-EF (r=0.891).
• In studies with UEAs (n=126), AI correlation was r=0.89 versus MD-EF's r=0.90.
• In unenhanced studies (n=116), AI correlation was r=0.87 versus MD-EF's r=0.84. Bland-Altman analysis showed a bias of 3.63 ± 7.40% for LVivoEF versus cMRI.

Conclusions:

• LVivoEF accurately quantifies LVEF from a standard apical four-chamber view without manual correction, validated against cMRI.
• The AI software shows potential to enhance and accelerate LVEF quantification in clinical practice.