Ultrasonic Elastography High-precision Displacement Tracking Method Research

Ultrasound elasticity imaging is a technique used to estimate the elastic properties of soft tissue. The information comes from displacement by tracking the ultrasonic signal, and, therefore, the high-precision displacement estimation method for elastography is very important. As a non-invasive, accurate and low-cost technique, ultrasound elstography has gradually replaced the manual palpation for early lesions of several organs in clinical diagnosis. Currently, improving the accuracy and speed of ultrasound speckle tracking is very important for researchers who are interested in ultrasound elatogrpahy application. In this paper, the following works have been done for these two issues:1) Improving computation speed:Recently, Dr. Jiang proposed a new sub-sampling displacement estimation method that can improve the accuracy of axial and lateral motion estimation, which exhibits performance advantages and a high flexibility over other methods. However, it is difficult to obtain real-time performance on CPU architecture. Within this approach, internal thread communication does not exist between the displacement estimation points. It is suitable for parallel computing on GPU. In the present study, it enables the efficient computation method on the GPU by the NVIDIA CUDA architecture. In contrast to the original method implemented by C Mex code, the GPU-based method achieved 76X acceleration while maintaining high accuracy sub-sampling displacement estimation.2) Improving the accuracy of displacement:Based on the consistency of the physical properties (i.e., elastic modulus represents hardness of tissue organization or represented by the strain profile) of the local area of human tissue and continuity of local displacement, this paper proposes a corresponding regularization method to remove noise and improve the accuracy of displacement motion tracking. More specifically, the proposed method is based on two-dimensional (2D) displacement estimation derived from the traditional displacement estimation method. The corresponding optimal solution was to meet conditions for minimizing the regularization problem, which is solved by the particle swarm algorithm. In this study, the consistent components of the elastic modulus of human tissue were physical constraints, represented by the above-mentioned mathematical regularization. Testing was done using computer-synthesized data and live data experiment, and its results showed that this method is able to effectively improve the accuracy of the lateral and axial estimated displacement and, eventually, obtained a higher quality axial and lateral strain image. Preliminary results show that the method, either from the concept or from the calculation methods, improved image quality for current clinical ultrasound elastography. As a post-processing tool, the proposed approach is very useful for improving displacement estimation of both directions.