Suppressing the HIFU interference in ultrasound guiding images with a diffusion-based deep learning model

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Artificial Intelligence

Background:

• Ultrasound-guided high-intensity focused ultrasound (HIFU) surgery requires simultaneous therapy and imaging.
• Real-time monitoring is challenging due to acoustic interference from HIFU sonication corrupting ultrasound images.

Purpose of the Study:

• To develop a deep learning (DL) solution to suppress HIFU interference in ultrasound imaging.
• To enable simultaneous HIFU therapy and real-time ultrasound imaging for enhanced surgical monitoring.

Main Methods:

• Proposed a diffusion implicit model (HIFU-Diff) to suppress HIFU interference.
• Treated interference suppression as a high-resolution image generation task via diffusion models.
• Validated the network using extensive ex-vivo and in-vivo experimental data.

Main Results:

• HIFU-Diff significantly outperformed notch filtering and 1D FUS-Net in reconstructing interference-free images.
• Demonstrated superior performance under strong interference and complex in-vivo conditions.
• Reconstructed images preserved acoustic characteristics, enabling quantitative attenuation imaging.

Conclusions:

• The HIFU-Diff model allows simultaneous HIFU therapy and ultrasound imaging for real-time treatment monitoring.
• This advancement can improve the safety and efficacy of non-invasive treatments.
• This is the first diffusion-based model developed for HIFU interference suppression.