This paper presents a robot-assisted system developed for spine surgery of percutaneous vertebroplasty. The system plays roles of assisting surgeon for puncture. Compared to navigation systems as well as conventional methods for percutaneous vertebroplasty, it is able to achieve better accuracy. With the help of the surgical planning system based on CT images, the surgeon could easily obtain the information of the target point, orientation, and depth of the surgical needle. Unlike other system based on vision, this robot system could achieve the target point with the desired angle without optical tracking system. A new method of coordinate transformation from the CT images to the robot is developed to cope with the orientation for the robot-assisted system.