Development of a Realistic Tissue Mimicking Phantom Using Finite Element Method

Abstract

In recent medical diagnosis, ultrasound strain imaging or elastography has been established as a useful technique to determine cancerous or abnormal tissues because a malignant tissue is stiffer than benign (normal) tissue. In elastography, the tissue strain is basically estimated from the gradient of tissue displacements and displacements of tissue are estimated from the time delays of gated pre- and post-compression echo signals. Most of the algorithms that are used to find the tissue displacement in elastography are one dimensional that means the displacement is measured only for axial direction which does not provide us with other directional deformation. Moreover, the direct strain imaging techniques are very computational costly that means it took lots of time to compute the data for imaging. To overcome those above discussed problem, this thesis introduces 2-D cross-correlation algorithm to compute time delay which could give us a more realistic reminiscence. In addition to that, the work flow of the strain imaging has been modified to make the imaging less computational costly. Also, to substantiate the above claim MATLAB tic toc function is used to determine each step simulation time of the modified work flow. In this thesis, the presented imaging technique is based on signal correlation. Correlation of signal is typically a measure of similarity of two signals as a function of the displacement of one relative to the other. At first, a 2D finite element tissue mimicking phantom model based on real tissue properties using ANSYS software having a tumor inclusion surrounded by soft tissue is developed. Then, the synthetic tissue was simulated for different inclusion tissue properties for a given displacement. After that, pre- and post-compression data was exported from the ANSYS model. Then, ultrasound simulation (using FIELD II) was done to generate pre- and postcompression RF signal for a more realistic simulation. After that, 2D cross correlation MATLAB function was used to estimate the time delay between pre- and post-compression RF signals. Using MATLAB surf tool, the estimated correlation coefficient was mapped. The map shows a promising result to distinguish between a normal and abnormal tissue. Also, the proposed algorithm offers less computation cost in terms of simulation time with a value of 222.072322 seconds in contrast with the simulation time of the conventional strain imaging having a value more than 222.093015 seconds. So, by applying the above imaging algorithm and procedure to a real world 3-D scenario, we could get a more sophisticated imaging technique which is also computation costly.