Artificial Intelligence-Based Left Ventricular Ejection Fraction by Medical Students for Mortality and Readmission...

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Point-of-care ultrasound (POCUS) is integral to modern medical diagnostics.
• Physicians across specialties utilize POCUS for enhanced decision-making.
• AI integration in POCUS offers potential for improved diagnostic accuracy.

Purpose of the Study:

• To evaluate the association between reduced left-ventricular ejection fraction (LVEF) detected by AI-assisted POCUS operated by medical students.
• To assess the 1-year composite outcome of mortality and cardiovascular readmission in patients with reduced LVEF.
• To determine the predictive value of AI-based LVEF assessment for patient outcomes.

Main Methods:

• Eight trained medical students performed POCUS using a hand-held ultrasound device (HUD) with an AI tool for LVEF evaluation.
• The study included 82 hospitalized cardiology patients from March 2019 to March 2020.
• AI-based LVEF was automatically assessed for all participants.

Main Results:

• 34 patients (41.5%) had AI-detected reduced LVEF (<50%).
• Reduced LVEF was associated with significantly higher 1-year composite outcomes (41.2% vs. 16.7%, p=0.014).
• Reduced LVEF independently predicted the composite outcome (HR 2.717, p=0.033) and was linked to longer hospital stays and adverse in-hospital events.

Conclusions:

• AI-assisted POCUS by medical students effectively identifies reduced systolic function.
• Reduced LVEF detected via AI-POCUS independently predicts 1-year mortality and cardiovascular readmission.
• AI utilization by novice users shows promise for risk stratification in hospitalized cardiac patients.