Аннотация:The article is concerned with developing effective methods for solving inverse problems of wave tomography. The underlying mathematical model accounts for such physical effects as ultrasound diffraction, absorption, refraction and multiple scattering of ultrasound waves. Reconstructinga tomographic image in wave tomography involves solving a coefficient inverse problem for the wave equation. The objective is to minimize the residual functional between the measured wave field and the wave field computed for some approximate acoustic parameters of the inspected object. The residual functional may have local minima. To solve such problems, a multistage iterative gradient-based method is proposed. The method uses limited signal spectrum at first stages, ensuring the convergence of approximate solutions obtained at each stage to the sought-for solution. The proposed method is demonstrated on model problems of breast ultrasound tomographic imaging for early-stage cancer diagnosis. The iterative method is computationally expensive. Computational efficiency of the proposed method on multi-core CPU and GPU computing platforms is discussed.