Development of high-performance scalable software for ultrasound tomography

This paper is dedicated to the development of high-performance scalable software for supercomputers to be used for solving the inverse problem of ultrasound tomography in medical applications. The problem is formulated as a nonlinear coefficient inverse problem for the wave equation. A program structure is proposed, which allows one to efficiently parallelize the computations on cluster systems. The code is optimized in a such way that the speed of the program is increased by an order of magnitude and a much higher scalability is provided. Many test runs of the program are performed on the Chebyshev and Lomonosov computing clusters installed at the Moscow State University.