Machine Learning-Enhanced Skull-Universal Acoustic Hologram for Efficient Transcranial Ultrasound Neuromodulation...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Acoustics

Background:

• Ultrasound neuromodulation (UNM) offers non-invasive, precise deep brain stimulation.
• The skull distorts ultrasound, requiring acoustic aberration correction for effective UNM.
• Current acoustic holograms are skull-specific, limiting scalability and requiring custom lenses.

Purpose of the Study:

• To develop a universal acoustic hologram for transcranial ultrasound neuromodulation.
• To enable efficient and scalable UNM across various skull types.
• To overcome the limitations of skull-specific acoustic lenses.

Main Methods:

• Developed a novel Skull-Universal Acoustic Hologram (SUAH) generation framework.
• Integrated physics-based acoustic holography with differentiable acoustic simulation.
• Employed a gradient accumulation technique for hologram optimization.
• Trained SUAH on a diverse range of rodent skull shapes.

Main Results:

• SUAH demonstrated remarkable generalizability and robustness across different skull types.
• SUAH outperformed skull-specific acoustic holograms (SSAH).
• Performance remained high even with smaller training datasets, surpassing individual skull-based training.

Conclusions:

• SUAH is a scalable, versatile, and accurate tool for transcranial ultrasound neuromodulation.
• This technology eliminates the need for multiple custom lenses, facilitating broader research.
• SUAH represents a significant advancement for ultrasound neuromodulation research.