Lujiang Liu
Shira L. Broschat
School of Elec. Eng. and Comput. Sci., Washington State Univ., Pullman, WA 99164-2752
In ultrasound mammography the objective is to image tumors about 4 or 5 mm in size. Thus the resolution of the reconstructed object profile is critical. Because of the strong refraction caused by fat lobes in the breast, simple pulse echo methods are inadequate, and a full inverse imaging technique is needed. The iterative Born method is an inverse technique that has been used in ultrasound imaging. However, the calculation cost is very high, and parallel computation is limited to the forward problem. In this work, a method is presented that improves the performance of the iterative Born approach, giving better resolution at less cost. The dependence between the incident source angle and the reconstructed object profile is explored, and an algorithm is presented that is realizable on a parallel computer for both forward and inverse calculations. Four single sources are evenly spaced around the object to be imaged and are used to reconstruct the object simultaneously; the results from the four separate sources are then averaged. The cost of this method is 20% less than the iterative Born method. In addition, this method is robust to noise since averaging helps to cancel noise. [Work supported by NSF.]