A GPU-Accelerated 3-D Coupled Subsample Estimation Algorithm for Volumetric Breast Strain Elastography.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Ultrasound Elastography

Background:

• Ultrasound breast strain elastography is crucial for detecting breast cancer.
• Previous 2-D algorithms limited 3-D strain analysis due to computational demands.
• Accelerating 3-D speckle tracking is essential for clinical feasibility.

Purpose of the Study:

• To extend a 2-D coupled subsample tracking algorithm to 3-D for ultrasound breast strain elastography.
• To investigate the use of graphic processing units (GPUs) for accelerating 3-D speckle tracking.
• To evaluate the performance of the GPU-accelerated 3-D algorithm against conventional methods.

Main Methods:

• Developed a 3-D coupled subsample tracking algorithm.
• Implemented a graphic processing unit (GPU) for computational acceleration.
• Validated the algorithm using a tissue-mimicking phantom and in vivo breast ultrasound data, comparing it to a conventional 3-D quadratic subsample estimation algorithm.

Main Results:

• The GPU implementation achieved high-quality strain data, evidenced by high correlation and contrast-to-noise ratio.
• The proposed 3-D subsample tracking algorithm demonstrated superior performance compared to the conventional 3-D quadratic subsample algorithm.
• Volumetric strain data was acquired rapidly (approx. 20 s per volume) using the GPU implementation.

Conclusions:

• The GPU-accelerated 3-D subsample estimation algorithm significantly enhances ultrasound breast strain elastography.
• This method provides high-quality volumetric strain data efficiently, overcoming previous computational limitations.
• The accelerated approach holds promise for faster and more accurate breast cancer diagnosis.