Compressed Sensing with Gaussian Sampling Kernel for Ultrasound Imaging

Area of Science:

• Medical Imaging
• Ultrasound Technology
• Signal Processing

Background:

• Compressed sensing (CS) is utilized in ultrasound imaging for data reduction and frame rate enhancement.
• Limited research exists on lateral undersampling strategies for conventional focused beamforming (CFB) within CS frameworks.

Purpose of the Study:

• To propose and evaluate a strategic lateral undersampling approach for channel data in ultrasound imaging using a Gaussian sampling scheme.
• To compare the proposed Gaussian sampling with uniform and 2-D random undersampling methods for CS recovery.

Main Methods:

• Developed a Gaussian sampling scheme for lateral undersampling of channel data, acquiring radiofrequency data from reduced receive elements.
• Compared the Gaussian scheme against uniform and 2-D random undersampling through simulations and experimental data.
• Investigated the impact of various lateral and axial undersampling rates on CS recovery.

Main Results:

• Gaussian-based channel data subsampling demonstrated superior CS recovery and image contrast compared to uniform undersampling.
• Despite discarding 90% of original data, the CS-recovered image contrast was comparable to the reference image.
• The proposed method strategically uses fewer active receive elements, significantly reducing data size.

Conclusions:

• The proposed Gaussian sampling scheme offers an effective strategy for lateral undersampling in CS-based ultrasound imaging.
• This approach significantly reduces data requirements and enhances image quality.
• It presents a promising option for developing affordable point-of-care ultrasound systems.