Machine Learning-Empowered Real-Time Acoustic Trapping: An Enabling Technique for Increasing MRI-Guided Microbubble...

Area of Science:

• Biomedical Engineering
• Acoustic Manipulation
• Medical Imaging

Background:

• Acoustic trapping uses ultrasound for non-invasive bioparticle manipulation.
• Magnetic resonance imaging (MRI) advances enhance acoustic interference sensing and drug carrier tracking.
• Improving MRI-guided microbubble (MB) accumulation in target microvessels is crucial for drug delivery.

Purpose of the Study:

• To develop a machine learning model for modulating transducer arrays in MRI-guided acoustic trapping.
• To address challenges in accurate acoustic trap generation due to complex ultrasound propagation.
• To enhance computational efficiency for real-time adjustments of acoustic traps.

Main Methods:

• A machine learning-based model was developed to predict time-of-flight (ToF) and pressure amplitude.
• The model modulates transducer arrays for precise control of acoustic interference.
• Model performance was validated using different transducer sizes and penetration depths.

Main Results:

• The model achieved low average prediction errors for ToF (-0.45 µs to 0.67 µs) and amplitude (-0.34% to 1.75%).
• Rapid prediction (<10 ms) demonstrated a four-order of magnitude improvement in computational efficiency over existing methods.
• Validation confirmed the model's adaptability and potential for future ultrasound treatments.

Conclusions:

• The proposed machine learning model significantly improves the accuracy and efficiency of MRI-guided acoustic trapping.
• This approach holds promise for enhanced drug carrier concentration and targeted therapies.
• The model's adaptability suggests broad applicability in advanced ultrasound-based medical interventions.