Advancing Single-Plane Wave Ultrasound Imaging With Implicit Multiangle Acoustic Synthesis via Deep Learning

Area of Science:

• Medical Ultrasound
• Image Processing
• Deep Learning

Background:

• Plane wave imaging (PWI) offers ultrafast capabilities for real-time physiological monitoring in medical ultrasound (US).
• Enhancing PWI image quality traditionally requires more plane waves (PWs), which reduces frame rates.
• Existing deep learning (DL) methods often use single-PW inputs, failing to capture steered wave information.

Purpose of the Study:

• To develop a DL architecture for PWI that implicitly integrates multiangle information.
• To improve PWI image quality and stability without compromising high frame rates.
• To overcome limitations of single-PW input methods in capturing steered wave data.

Main Methods:

• Developed a DL architecture to generate and combine virtual steered PWs from a single input view.
• Utilized an attention mechanism to merge virtual PWs with actual single-PW data.
• Employed implicit multiangle acoustic synthesis for high-quality image reconstruction.

Main Results:

• Achieved high-quality imaging output comparable to extensive multiangle compounding.
• Demonstrated superior performance over traditional single-PW strategies in simulations, phantoms, and in vivo targets.
• Showcased enhanced stability, reliability, and robustness in imaging outcomes, preserving speckle patterns and diagnostic details.

Conclusions:

• The proposed method advances PWI technology by enabling high-quality imaging without sacrificing speed.
• Implicit multiangle acoustic synthesis offers a promising solution for real-time US imaging challenges.
• The DL approach effectively restores crucial details for in vivo diagnostic imaging.