An image-guided robot for needle insertion in small animal. Accurate needle positioning using visual servoing.

The development of new medical therapies often require experiments on small animals. In order to improve the medical protocol during treatments with needles, we propose a new robotic needle insertion system using CT-scan imaging and visual servoing. The biologist defines the skin entry point and the target to be reached in the CT-image. The needle target is then expressed in the robot frame thanks to a 3D registration between the 3D structured light reconstruction of the animal bed and the 3D model obtained from the CT-scan data. Finally the needle pointing is performed thanks to a 2D visual servoing. This approach is independent of the needle model and allows a high positioning accuracy: less than 0.5 pixel for the tip of the needle and less than 10^-3 rad for the direction. Experiments show promising results: a needle with a diameter of 0.72 mm can be inserted through two holes with diameters of 2 mm.