Artificial Intelligence for Detection of Prognostically Significant Left Ventricular Dysfunction From...

Area of Science:

• Cardiology
• Artificial Intelligence in Medicine
• Medical Imaging Analysis

Background:

• Echocardiography is crucial for assessing left ventricular (LV) function.
• Identifying LV dysfunction can be challenging, especially with preserved ejection fraction (EF).
• Current methods may struggle with incomplete or ambiguous echocardiographic data.

Purpose of the Study:

• To evaluate the performance of an artificial intelligence LV dysfunction (AI-LVD) identification algorithm.
• To determine the algorithm's ability to identify LV dysfunction from routine echocardiographic measurements.
• To assess the prognostic value of AI-LVD, particularly in cases with preserved EF.

Main Methods:

• Trained AI-LVD on large echocardiographic datasets (imputation and training cohorts).
• Validated AI-LVD in a separate cohort of 81,509 patients with diverse EF categories.
• Analyzed AI-LVD's correlation with all-cause mortality and its performance with missing data.

Main Results:

• AI-LVD demonstrated sex-specific outputs correlating with LV dysfunction severity and mortality risk.
• The algorithm showed significant prognostic capacity in patients with preserved EF.
• Higher AI-LVD deciles were associated with substantially increased 5-year all-cause mortality in both men and women.

Conclusions:

• A novel AI-LVD algorithm effectively identifies prognostically significant LV dysfunction using only echocardiographic data.
• The AI-LVD is reliable even when key echocardiographic parameters are missing or EF is preserved.
• This AI tool can be integrated into routine echocardiography for real-time reporting and risk stratification.