Related Experiment Video
Updated: Dec 22, 2025
05:57
Blood Flow Imaging with Ultrafast Doppler
Published on: October 14, 2020
8.2K
Real time SVD-based clutter filtering using randomized singular value decomposition and spatial downsampling for
U-Wai Lok1, Pengfei Song1, Joshua D Trzasko1
1Department of Radiology, Mayo Clinic College of Medicine and Science, Rochester, MN, USA.
Ultrasonics
|May 1, 2020
Summary
Randomized singular value decomposition (rSVD) significantly accelerates clutter filtering for real-time ultrasound micro-vessel imaging. This method achieves high Blood-to-Clutter Ratio (BCR) and frame rates, crucial for tumor diagnosis.
Area of Science:
- Medical Imaging
- Ultrasound Technology
- Signal Processing
Background:
- Singular Value Decomposition (SVD) offers robust clutter filtering in ultrasound but faces computational challenges for real-time applications.
- Conventional high-pass filters are less effective than SVD for tissue clutter rejection.
- Real-time SVD-based clutter filtering requires high frame rates (≥10-15 Hz) for effective micro-vessel imaging.
Purpose of the Study:
- To implement and evaluate a previously proposed acceleration method using randomized SVD (rSVD) and spatial downsampling on an ultrasound scanner.
- To optimize imaging and processing parameters for real-time micro-vessel imaging using the rSVD method.
- To assess the performance of the rSVD-based clutter filter in terms of Blood-to-Clutter Ratio (BCR) and frame rate.
Main Methods:
- Implementation of the rSVD clutter filtering and randomized spatial downsampling method on a Verasonics ultrasound scanner with a multi-core CPU.
- Evaluation of parameter selections, including block size and ensemble size, to achieve real-time processing.
- Real-time demonstration using a 12-core CPU with a downsampling factor of 12 and 12 threads.
Main Results:
- The rSVD-based clutter filter achieved processing times under 30 ms with BCRs exceeding 20 dB when using specific parameters (block size 30x30, ensemble size 45, rank 26).
- A micro-vessel imaging frame rate of approximately 22 Hz was achieved with optimized parameters (block size 20x20, ensemble size 45, rank 26).
- The method demonstrated its capability for real-time 2D scanning for tumor microvessel visualization.
Conclusions:
- The implemented rSVD acceleration method enables real-time micro-vessel imaging with high clutter rejection and frame rates.
- This approach overcomes the computational limitations of traditional SVD for ultrasound applications.
- The technology holds potential for improved diagnosis of tumor microvasculature through enhanced 2D visualization.
