A GPU-Parallelized Eigen-Based Clutter Filter Framework for Ultrasound Color Flow Imaging.

Area of Science:

• Medical Imaging
• Signal Processing
• Computational Ultrasound

Background:

• Eigen-filters improve color flow imaging (CFI) sensitivity by adapting to clutter statistics.
• Clinical adoption is limited by the high computational cost of eigendecompositions.

Purpose of the Study:

• To describe a real-time computing framework for eigen-based clutter filtering using a single-instruction, multiple data (SIMD) approach on a graphical processing unit (GPU).
• To demonstrate the feasibility and performance of GPU-accelerated eigen-filtering for enhanced CFI.
• To address the computational challenges hindering the clinical use of advanced clutter filtering techniques.

Main Methods:

• Developed a real-time computing framework leveraging single-instruction, multiple data (SIMD) parallel processing on a GPU.
• Employed the single-ensemble-based eigen-filtering approach (Hankel singular value decomposition) for algorithmic compatibility with GPU SIMD computing.
• Explained key algebraic principles and SIMD algorithm implementation on the GPU.

Main Results:

• Achieved real-time video-range throughput (24 frames/s) for CFI frames up to 5 cm scan depth with a slow-time ensemble length of 16 samples on a single GPU (GTX Titan X).
• Demonstrated enhanced flow detection sensitivity in vivo compared to non-adaptive polynomial regression clutter filtering.
• Validated the framework's real-time efficacy and performance across varying scan depths and ensemble lengths.

Conclusions:

• GPU-enabled eigen-based clutter filtering offers a viable solution for real-time processing in CFI.
• The developed framework significantly improves flow detection sensitivity and can enhance clinical diagnostic capabilities.
• This approach overcomes computational barriers, paving the way for broader clinical adoption of advanced CFI techniques.