Artificial Intelligence-Guided Lung Ultrasound by Nonexperts

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Point-of-Care Ultrasound

Background:

• Lung ultrasound (LUS) is crucial for diagnosing dyspnea but requires specialized expertise.
• Artificial intelligence (AI) has shown promise in guiding novice users for cardiac ultrasound image acquisition.
• The potential of AI to enhance LUS image acquisition for broader clinical application warrants investigation.

Purpose of the Study:

• To assess the efficacy of AI in enabling trained healthcare professionals (THCPs) to obtain diagnostic-quality LUS images.
• To validate the performance of an AI-guided LUS system against expert-acquired images.

Main Methods:

• A multicenter diagnostic validation study involved 176 participants with shortness of breath.
• Trained healthcare professionals (THCPs) used AI software (Lung Guidance AI) for LUS image acquisition.
• AI software utilized deep learning algorithms to guide image capture and B-line annotation following an 8-zone protocol.
• Images were reviewed by a panel of masked LUS experts for diagnostic quality assessment.

Main Results:

• 98.3% of LUS examinations acquired by THCPs with AI assistance were deemed of diagnostic quality.
• No significant difference in diagnostic quality was observed between AI-assisted THCPs and expert sonographers.
• The AI system automatically captured images of diagnostic quality using a standardized 8-zone LUS protocol.

Conclusions:

• AI-assisted LUS acquisition by THCPs can achieve diagnostic quality comparable to expert sonographers.
• This AI technology has the potential to significantly increase access to LUS diagnostics.
• Implementation of AI guidance can help overcome personnel shortages in expert LUS interpretation, particularly in underserved areas.