Complex Residual Attention U-Net for Fast Ultrasound Imaging from a Single Plane-Wave Equivalent to Diverging Wave...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Ultrasound
• Signal Processing

Background:

• Plane wave imaging offers high frame rates but suffers from noise, speckle, and artifacts, degrading image quality.
• Reconstruction of complex ultrasound data is crucial for accurate image interpretation.
• Existing methods may not fully address the limitations of single-acquisition plane wave imaging.

Purpose of the Study:

• To develop an advanced deep learning model for reconstructing high-quality complex ultrasound data from single plane wave insonification.
• To improve image quality and resolution by mitigating noise and artifacts inherent in plane wave imaging.
• To match the performance of more complex coherent compounding imaging techniques.

Main Methods:

• Proposed an attention-based complex convolutional residual U-Net architecture.
• Incorporated an attention mechanism within the complex domain and utilized complex convolution.
• Trained the network on simulated phased array data and validated with in vitro and in vivo ultrasound datasets.

Main Results:

• The proposed method successfully reconstructed improved in-phase/quadrature complex data.
• Attention mechanism focused on critical data aspects, effectively separating regions of interest from background noise.
• Achieved enhanced ultrasound image quality compared to standard methods.

Conclusions:

• The attention-based complex U-Net effectively reconstructs high-quality ultrasound data from plane wave imaging.
• This approach offers a promising solution for improving image quality and diagnostic accuracy in ultrasound.
• The method demonstrates potential for real-time applications requiring high frame rates and excellent image fidelity.