Limited echocardiogram acquisition by novice clinicians aided with deep learning: A randomized controlled trial

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Cardiology

Background:

• Global sonographer shortage limits timely ultrasound diagnostics.
• Deep learning (DL) offers potential to improve image acquisition by non-specialists.
• Limited access to crucial diagnostic imaging in resource-limited settings.

Purpose of the Study:

• To evaluate if DL-enabled devices enhance the acquisition of multi-view limited echocardiograms by internal medicine residents without prior cardiac ultrasound training.
• To assess the impact of DL on scan time and image quality.

Main Methods:

• Single-center randomized controlled trial with 38 internal medicine residents over two weeks.
• Participants used portable ultrasound devices with or without DL real-time guidance.
• Primary outcome: total acquisition time; Secondary outcomes: image quality and participant attitudes.

Main Results:

• DL-equipped devices significantly reduced total scan times (152s vs. 266s, P < 0.001).
• DL group achieved superior image quality (modified RACE scores 15 vs. 11, P = 0.034).
• Performance improvements were most notable in challenging views.

Conclusions:

• Deep learning-enhanced ultrasound devices significantly improve echocardiogram acquisition speed and image quality for novice users.
• DL-enhanced ultrasound may address diagnostic imaging gaps by enabling non-specialists to acquire clinically useful cardiac images.