Harnessing chemically crosslinked microbubble clusters using deep learning for ultrasound contrast imaging.

Area of Science:

• Biomedical Engineering
• Acoustic Physics
• Machine Learning

Background:

• Microbubble clusters (CCMCs) are advanced contrast agents with unique acoustic properties.
• Distinguishing CCMCs from individual microbubbles (MBs) is crucial for targeted ultrasound applications.
• Machine learning (ML) offers novel methods for analyzing complex acoustic data.

Purpose of the Study:

• To investigate and isolate the distinctive acoustic properties of CCMCs.
• To apply ML, specifically autoencoder-based anomaly detection, for CCMC acoustic analysis.
• To enhance the detectability and localization of contrast agents in ultrasound imaging.

Main Methods:

• CCMCs synthesized using copper-free click chemistry.
• Acoustic analysis of CCMCs and individual MBs using a clinical transducer.
• Radiofrequency data processed for ML model training and testing (anomaly detection).

Main Results:

• The anomaly detection model successfully identified unique acoustic signatures of CCMCs.
• Frequency analysis revealed higher amplitude and energy in CCMC acoustic signals, indicating coalescence.
• Control experiments confirmed the model's specificity in distinguishing clustered from non-clustered MBs.

Conclusions:

• ML-based anomaly detection is feasible for identifying CCMC acoustic characteristics.
• CCMCs exhibit elevated acoustic amplitudes, beneficial for contrast agent detection.
• This approach holds promise for improved ultrasound imaging, particularly super-resolution applications.