FPGA Implementation of the Coupled Filtering Method and the Affine Warping Method

Area of Science:

• Medical imaging
• Biomedical engineering
• Ultrasound technology

Background:

• Speckle tracking is crucial for ultrasound-based tissue elasticity analysis.
• Feature-motion decorrelation and high computational complexity limit current methods, even with GPU acceleration.

Purpose of the Study:

• To develop and evaluate Field-Programmable Gate Array (FPGA) implementations of coupled filtering and affine warping methods for enhanced ultrasound speckle tracking.
• To address the computational bottlenecks of existing high-deformation tissue analysis techniques.

Main Methods:

• Reformulation of algorithms for efficient pipelining on FPGAs.
• Development of a fast, memory-saving image warping approach tailored for FPGA architecture.
• Implementation and testing on a Xilinx Virtex-7 FPGA platform.

Main Results:

• FPGA-based implementations significantly accelerate strain estimation in ultrasound image analysis.
• Achieved speedup of at least 13 times compared to a high-performance GPU (NVIDIA GeForce GTX 580).
• Demonstrated the effectiveness of FPGAs in handling complex image processing for speckle tracking.

Conclusions:

• FPGA acceleration provides a viable solution for computationally intensive ultrasound speckle tracking methods.
• The proposed FPGA approach enhances the efficiency of tissue elasticity analysis, particularly for large deformations.
• This work paves the way for real-time, high-performance ultrasound elastography.