In our approach, we first specify a 3D model of the image structure by segmenting a CT dataset into the respective tissues followed by assigning the acoustic properties (velocity, impedance, scattering mean and standard deviation, damping factor and packing factor). Given that model, we simulate the ray propagation, beam forming, and finally the backscattering. Due to the inhomogenities of tissue, different physical models for ultrasound simulation are required: Rayleigh scattering is applied for homogenous regions and ray tracing techniques handle abrupt changes in acoustic impedance on tissue boundaries. The latter leads to different phenomena like refraction (Snell's law), reflection and transmission (Fresnel equation). The gradients needed for these methods are precomputed for each model using a central-difference method with multiple neighbours. Absorption is calculated by the Beer-Lambert law.