I have implemented a TTI RTM by solving the coupled equations with a hybrid FD and PS method. Numerical tests show that a stable TTI RTM is achievable by doing a selective anisotropic parameter equating in the model to reduce the difference of epsilon and delta in areas of rapid changes in the symmetry axes. A threshold value needs to be empirically selected for the equating using a combination of epsilon, delta and theta values. It is clear that TTI RTM has the ability to produce a more accurate image than isotropic RTM, especially in areas with anisotropy and strong variations of dip angle. For 3D, spatial variation of the symmetry axis is much more difficult to handle than 2D cases. This is a topic of future research. Also I will focus on how to reduce memory and computational costs for 3D TTI RTM.