Identifying Unique Acoustic Signatures from Chemically-Crosslinked Microbubble Clusters Using Deep Learning.

Area of Science:

• Biomedical Engineering
• Acoustics
• Materials Science

Background:

• Ultrasound contrast agents (UCAs) are vital for enhanced ultrasound imaging and drug delivery.
• Current UCAs require improved detection algorithms for faster and more accurate performance.
• Chemically Cross-linked Microbubble Clusters (CCMCs) represent a novel class of lipid-based UCAs.

Purpose of the Study:

• To demonstrate the unique acoustic response of CCMCs compared to individual UCAs.
• To investigate the potential of deep learning for distinguishing CCMCs based on their acoustic signatures.
• To explore CCMCs for novel contrast agent detection techniques.

Main Methods:

• Acoustic characterization of CCMCs and individual UCAs using broadband hydrophone and Verasonics Vantage 256 with a clinical transducer.
• Development and training of a simple artificial neural network (ANN) to classify raw 1D RF ultrasound data.
• Classification of data from CCMC and individual UCA populations.

Main Results:

• The ANN achieved 93.8% accuracy in classifying CCMCs using broadband hydrophone data.
• The ANN achieved 90% accuracy in classifying CCMCs using Verasonics clinical transducer data.
• Results indicate distinct acoustic responses for CCMCs.

Conclusions:

• CCMCs exhibit unique acoustic properties differentiating them from individual UCAs.
• Deep learning algorithms can effectively classify CCMCs based on their acoustic signatures.
• CCMCs hold promise for developing advanced contrast agent detection methods in ultrasound.