Locally optimized correlation-guided Bayesian adaptive regularization for ultrasound strain imaging.

Area of Science:

• Medical imaging
• Biomedical engineering
• Ultrasound technology

Background:

• Ultrasound strain imaging estimates tissue elasticity using radio-frequency (RF) echo signals for diagnostic value.
• Accurate displacement estimation from RF signals is critical for high-quality strain tensor images.
• Regularization improves accuracy and precision in displacement estimation frameworks.

Purpose of the Study:

• To propose an adaptive Bayesian regularization scheme for enhanced ultrasound strain imaging.
• To improve the accuracy and precision of displacement and strain estimation.
• To validate the method in quasi-static, cardiac, and in vivo applications.

Main Methods:

• Developed an adaptive Bayesian regularization scheme integrated into a 2D multi-level block matching algorithm.
• Algorithm dynamically adjusts iteration numbers based on normalized cross-correlation (NCC) quality and RF signal similarity.
• Validated using uniform and inclusion phantoms, canine cardiac deformation models, and in vivo murine heart data.

Main Results:

• Adaptive Bayesian regularization significantly reduced lateral strain error compared to fixed iteration methods (e.g., 27.51% vs. 104.49% at 1.5% strain).
• Improved contrast-to-noise ratios (CNR_e) for enhanced lesion detectability in inclusion phantoms (e.g., 7.42 dB vs. -0.31 dB lateral CNR_e at 1.5% strain).
• Demonstrated improved myocardial strain images in cardiac models and successful in vivo feasibility in murine hearts.

Conclusions:

• The proposed adaptive Bayesian regularization method enhances the robustness of ultrasound strain imaging.
• This technique offers significant improvements in accuracy and lesion detectability for clinical and preclinical applications.
• The adaptive approach provides a more reliable tool for tissue elasticity assessment using ultrasound.