Current Advances in Computational Lung Ultrasound Imaging: A Review.

Area of Science:

• Biomedical imaging
• Medical ultrasonography
• Lung imaging analysis

Background:

• Ultrasonography is a key non-ionizing, portable diagnostic tool in biomedical imaging.
• Lung ultrasound (LUS) examination leverages unique characteristics and artifacts for disease detection.
• Medical ultrasound image processing is crucial for accurate LUS interpretation.

Approach:

• Categorizes US image computing into model-based and data-driven methods.
• Discusses inverse problem-based techniques for US image enhancement (despeckling, deconvolution, artifact detection).
• Highlights deep/machine learning applications in supervised, weakly supervised, and unsupervised LUS image analysis.

Key Points:

• LUS imaging characteristics and artifact identification are central to disease diagnosis.
• Model-based methods address image quality issues like noise and blur.
• Data-driven methods, particularly machine learning, offer advanced analytical capabilities for LUS data.

Conclusions:

• The review provides a comprehensive overview of state-of-the-art medical US image processing for lung examinations.
• Both traditional model-based and modern data-driven approaches are vital for advancing LUS diagnostics.
• Understanding LUS artifacts and advanced image processing techniques improves diagnostic accuracy and utility.