Utility of a Deep-Learning Algorithm to Guide Novices to Acquire Echocardiograms for Limited Diagnostic Use.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Cardiovascular Ultrasound

Background:

• Artificial intelligence (AI) is increasingly used for medical image analysis.
• AI-guided acquisition of ultrasonography images, particularly echocardiograms, is an emerging field.
• A novel deep-learning (DL) algorithm has been developed to assist in obtaining diagnostic transthoracic echocardiographic images.

Purpose of the Study:

• To evaluate if novice users, trained with DL-based software, can achieve diagnostic quality in 10-view transthoracic echocardiographic studies.
• To assess the efficacy of AI in guiding inexperienced operators for cardiac ultrasound acquisition.

Main Methods:

• A prospective, multicenter diagnostic study involving 8 nurses with no prior echocardiography experience.
• Nurses were trained using a DL algorithm and scanned 30 patients each.
• Paired echocardiograms were obtained: one by a nurse with AI guidance, and one by a sonographer without AI guidance, with independent expert evaluation.

Main Results:

• The DL algorithm enabled nurses to produce diagnostic quality studies for left ventricular size and function (98.8%) and pericardial effusion (98.8%).
• Diagnostic quality for right ventricular size was achieved in 92.5% of cases.
• Comparisons showed no significant differences between nurse-acquired scans (with AI) and sonographer-acquired scans for most secondary endpoints.

Conclusions:

• The DL algorithm empowers novices to acquire diagnostic transthoracic echocardiograms for key cardiac assessments.
• This technology expands the accessibility of echocardiography to resource-limited settings and situations requiring immediate cardiac evaluation.
• AI-guided ultrasound acquisition holds significant potential for democratizing cardiac imaging expertise.