Planned delivery of focused therapy is adversely affected by internal body motion, such as from breathing, which could be mitigated, if tracked accurately in real-time. By extending an algorithm for superficial vein tracking , we have recently presented a robust real-time motion tracking method for 2D ultrasound image sequences of the liver . The method leverages elliptic and template-based models of vessels in the liver, coupled with a robust optic-flow framework resulting in iterative tracking. Potential drifts are corrected when the breathing phase is close to that of the initial reference frame, detected by comparing the appearance of tracked feature regions. Our method runs on a standard computer in real-time with latencies of 20-70 ms. It is implemented as a mobile platform to be connected from the display output of any ultrasound machine in clinics. Performance of our technique has been evaluated through the MICCAI Challenge on Liver Ultrasound Tracking (CLUST) 2015. This benchmark consisted of 64 2D ultrasound sequences (test-set with an average resolution of 447x552 px and an average length of 3761 frames at ~15 fps) collected from four institutes. For evaluation, up to 4 landmarks were selected in several frames by 3 different observers. Our method resulted in 1.09 (1.75) mm mean(std) error with 95th percentile at 2.42 mm, while the three annotators mean difference from their averaged consensus location was 0.5 mm with 95th percentile at 1 mm. Our system can be easily incorporated into the treatment room for real-time tracking of liver motion in 2D. Extension to 3D will be investigated next.