Wavelet Conditioning for Axial Strain Time Constant Estimation

Abstract

Ultrasound elastography has gained considerable consideration as a non-invasive imaging modality over the past ten years. Ultrasound elastography maps the strains experienced by a tissue under a small compression with the intent of acquiring diagnostic information about diseases. Elastographic data acquired from function-specific imaging mode can be utilized to distinguish tissue stiffness when subjected to a small external mechanical force. Poroelastography is a new elastographic technique that is used to image the temporal behavior of tissues subjected to a small uni-axial mechanical force. From the temporal data, poroelastographic imaging makes use of curve fitting techniques to estimate temporally-related elastographic parameters such as the Axial Strain Time Constant (ASTC). Curve fitting techniques can be directly applied to ultrasound data obtained from the soft tissue subjected to a constant compression. However, experimental poroelastographic data are inherently noisy due to long, oftentimes hand-held, data acquisition. The presence of this noise can affect the accuracy of the ASTC estimation. To estimate the ASTC with high accuracy, it is essential to reduce the noise that affects the data before applying the curve-fitting techniques. This work focuses on a method to denoise ultrasound poroelastographic data, which uses wavelet thresholding in order to improve the robustness of the curve fitting technique and increase the accuracy of the estimated ASTC. Wavelet thresholding is achieved through evaluating the wavelet coefficients and choosing the threshold optimally. Threshold value determines the efficiency of the denoising operation. In this work, we are using wavelet filtering along with the wavelet hard thresholding technique. Simulations are used to test the proposed denoising methods. Experiments were used to demonstrate the applicability of the proposed technique to experimental data. Simulation results show that the use of the wavelet-based denoising method prior computation of the ASTC using curve fitting can lead to ASTC estimates that are significantly more accurate than those obtained when the denoising method is not used.