Accelerated Singular Value-Based Ultrasound Blood Flow Clutter Filtering With Randomized Singular Value Decomposition...

Area of Science:

• Medical imaging
• Ultrasound technology
• Signal processing

Background:

• Singular value decomposition (SVD) enhances clutter rejection in ultrafast plane wave microvessel imaging.
• SVD is crucial for advanced applications like functional ultrasound and super-resolution imaging.
• The computational demands of SVD hinder its clinical application.

Purpose of the Study:

• To accelerate SVD-based clutter filtering for ultrafast ultrasound imaging.
• To introduce randomized SVD (rSVD) and randomized spatial downsampling as efficient alternatives.
• To evaluate the performance and computational benefits of these accelerated methods.

Main Methods:

• Implemented randomized SVD (rSVD) to approximate singular values, reducing computational load.
• Utilized randomized spatial downsampling to create smaller matrices for parallel processing.
• Validated methods using in vitro blood flow phantoms and in vivo rabbit kidney perfusion models.

Main Results:

• Proposed methods achieved significant computational acceleration (up to 6x combined) with minimal performance compromise (e.g., <3-dB reduction in blood-to-clutter ratio).
• rSVD demonstrated comparable clutter rejection to full SVD in vivo.
• Random downsampling, even at 60x, maintained artifact-free imaging with >10 dB blood-to-clutter ratio.

Conclusions:

• Randomized SVD and spatial downsampling effectively accelerate SVD clutter filtering for ultrafast microvessel imaging.
• These methods enable real-time imaging (approx. 40ms processing time).
• The findings support broader clinical translation and research in ultrafast microvessel imaging.