RESEARCHERS based in the USA have demonstrated that AI can help novice users of lung ultrasound (LUS) acquire diagnostic-quality images. The study has highlighted AI’s potential to make lung ultrasound imaging more accessible, particularly in resource-limited settings.
The study involved trained healthcare professionals, including medical assistants, respiratory therapists, and nurses, who were guided by AI software to perform lung ultrasounds on patients presenting with shortness of breath. Each patient underwent two examinations: one conducted by non-experts with AI assistance and another by expert sonographers without AI. A panel of blinded reviewers assessed the quality of the images obtained.
Results showed that 98.3% of the images captured by non-experts using AI met diagnostic standards, with no significant difference compared to the quality of expert-acquired images. Notably, in the left anterior inferior lung zone, where the presence of the heart often complicates imaging, AI-assisted users outperformed experts in capturing higher-quality clips.
“These findings underscore AI’s ability to close the skills gap in LUS acquisition, enabling confident use by operators with limited training,” researchers explained.
The average time to complete an AI-assisted LUS was 16.5 minutes, with physicians completing scans faster than non-physicians. This efficiency, coupled with high image quality, suggests AI-guided LUS could benefit both low-resource settings and advanced healthcare environments.
They highlighted potential applications, such as using AI-guided LUS for community health workers to screen for lung conditions like pneumonia and pulmonary oedema. The images could then be reviewed remotely by physicians, particularly useful when combined with portable ultrasound devices.
Future research will explore integrating AI-guided imaging with algorithms for detecting lung abnormalities, aiming to evaluate the technology’s effectiveness across diverse clinical settings.
“These developments could transform lung ultrasound from a specialised skill to an accessible diagnostic tool, improving patient care globally,” the team concluded.
Victoria Antoniou, EMJ
Reference
Baloescu C et al. Artificial intelligence–guided lung ultrasound by nonexperts. JAMA Cardiol. 2025;DOI:10.1001/jamacardio.2024.4991.
