Experimental investigation of computed tomography sound velocity reconstruction using incomplete data.

An approach for reconstructing the sound velocity distribution in the breast was previously proposed and verified by simulations, and the present study investigated the approach experimentally. The experimental setup comprised a 5-MHz, 128-channel linear array, a programmable digital array system, a phantom containing objects with differing physical properties, and a computer. The array system was used to collect channel data for simultaneous B-mode image formation and limited-angle tomographic sound velocity reconstruction. The phantom was constructed from materials mimicking the following tissues in the breast: glandular tissue, fat, cysts, high-attenuation tumors, and irregular tumors. The sound velocities in these materials matched those in the corresponding real tissues. The imaging setup is similar to that of x-ray mammography, in which a linear array is placed at the top of the breast and a metal plate is placed at the bottom for reflecting sound waves. Thus, both B-mode images and the sound velocity distribution can be acquired using the same setup. An algorithm based on a convex programming formulation was used to reconstruct the sound velocity images. By scanning the phantom at different positions, nine cases were evaluated. In each of the nine cases, the image object comprised a background (glandular tissue) and one or three regions of interest (fat, tumor, or cyst). The sound velocity was accurately estimated in the nine cases evaluated, with sound velocity errors being less than 5 m/s in 8 of 11 regions of interest. Thus, obtaining the sound velocity distribution is feasible with a B-mode imaging setup using linear arrays. Knowledge of the sound velocity distribution in the breast can be used to complement B-mode imaging and to enhance the detection of breast cancer.