Artificial intelligence for lung ultrasound interpretation: a systematic review

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pulmonology

Background:

• Lung ultrasound (LUS) is a valuable bedside diagnostic tool.
• Interpreting LUS images presents challenges due to a lack of expert training.
• Artificial Intelligence (AI) offers potential solutions for LUS interpretation.

Purpose of the Study:

• To systematically review and analyze recent AI-based tools for LUS interpretation.
• To summarize advances in AI for identifying lung artifacts, structures, and pathologies.
• To assess the current state of AI in supporting LUS diagnostics.

Main Methods:

• Systematic literature search of Web of Science, IEEE Xplore, and PubMed (2015-Nov 2025).
• Included studies employed AI for LUS image analysis.
• Risk of bias assessed using PROBAST+AI.

Main Results:

• Twenty-four studies were included, utilizing segmentation, object detection, and saliency maps.
• All studies employed Convolutional Neural Network (CNN) architectures.
• Significant heterogeneity in objectives, metrics, and risk-of-bias concerns (participants, analysis) were noted.

Conclusions:

• AI systems for LUS interpretation have high potential but face challenges.
• Current AI tools exhibit heterogeneity in design and evaluation.
• Standardized protocols and clinical environment-specific solutions are needed for practical implementation.