Two-dimensional least-squares estimation for motion tracking in ultrasound elastography

Area of Science:

• Medical Imaging
• Signal Processing
• Computer Vision

Background:

• Accurate motion tracking in ultrasound imaging is crucial for diagnosis.
• Traditional methods like Sum of Absolute Differences (SAD) often require image interpolation, increasing computational cost and potential errors.

Purpose of the Study:

• To propose a novel analytical method for 2-D translation estimation using a priori signal models.
• To develop a Least Squares Estimator (LSE) for precise phase adjustment and translation estimation.
• To integrate the LSE into a block matching algorithm for enhanced motion tracking in ultrasound images.

Main Methods:

• Definition of two analytical signals using multidimensional Hilbert transform to establish linear phase relationships with translations.
• Development of a Least Squares Estimator (LSE) to fit measured complex signal phases to theoretical forms, providing an analytical solution.
• Implementation of the LSE within a block matching framework for motion tracking and comparison with the Sum of Absolute Differences (SAD) method.

Main Results:

• The proposed LSE method provides an analytical solution for 2-D translation estimation without requiring image interpolation.
• For images at original resolution, the LSE method demonstrates significantly higher accuracy compared to the SAD cost function.
• Achieving comparable results to SAD on images with five times higher resolution, the LSE method is ten times faster.

Conclusions:

• The LSE-based method offers a more accurate and computationally efficient approach to 2-D translation estimation in ultrasound imaging.
• Eliminating the need for image interpolation, this method preserves image integrity and improves tracking performance.
• This technique presents a valuable advancement for real-time motion tracking applications in medical ultrasound.