Improving Lateral Resolution in 3-D Imaging With Micro-beamforming Through Adaptive Beamforming by Deep Learning.

Area of Science:

• Medical Imaging
• Ultrasound Technology
• Artificial Intelligence in Medicine

Background:

• Miniature ultrasound probes are crucial for in-body imaging, demanding high frame rates and volumetric capabilities.
• Achieving good lateral resolution with small apertures is challenging, especially when micro-beamforming reduces focusing.
• Existing beamforming methods offer insufficient resolution improvement or are computationally prohibitive.

Purpose of the Study:

• To develop an advanced beamforming technique for miniature ultrasound probes.
• To improve lateral resolution and overall image quality in volumetric ultrasound imaging.
• To create a practical and computationally efficient solution for in-body ultrasound applications.

Main Methods:

• Proposed adaptive beamforming by deep learning (ABLE) utilizing training targets from large aperture arrays.
• Modified ABLE to enhance its receptive field across multiple voxels for improved spatial accuracy.
• Trained the deep learning network using only in-silica data for practical implementation.

Main Results:

• Quantitatively and qualitatively demonstrated improved lateral resolution compared to conventional beamformers.
• Achieved superior image quality over delay-and-sum, coherence factor, filtered-delay-multiplication-and-sum, and Eigen-based minimum variance methods.
• Confirmed that in-silica data is sufficient for network training, ensuring ease of implementation.

Conclusions:

• Adaptive beamforming by deep learning (ABLE) offers a significant advancement for miniature ultrasound imaging.
• The proposed method effectively compensates for focusing reduction in micro-beamforming, enhancing lateral resolution.
• ABLE provides a practical and implementable solution for improving in-body volumetric ultrasound image quality.