Many types of research on robot-assisted minimally invasive surgery (RMIS) have been conducted, and its use in actual surgery is increasing. However, prior image information regarding the surgical target is required to generate a path for the surgical tool for RMIS. The image coordinate, target’s coordinate, and robot coordinate should be aligned through registration. However, errors are bound to occur during the image acquisition and registration. As the image acquisition time and registration time increase, the error between the patient and the coordinate information at the time of the actual operation increases due to the movement of the patient. To minimize these errors, this study proposes a registration-free approach to MIS without the preoperative phase in which a robot is used to directly contact the target to obtain a point cloud and reconstruct the shape information of the target. Using the position-based impedance and constraint controls for the remote center of motion (RCM) of the robot for MIS, the position information of the target can be acquired in the form of a point cloud without damage. Further, by converting the obtained point cloud into a mesh form using the Point2Mesh deep learning algorithm, it is possible to reconstruct the area where the position information is insufficient because there is no contact among the target areas. The process could obtain the coordinates within 3 min for the phantom. After a deep learning process of about 10 min, a surgical path using a robot could be generated. The reconstruction accuracy showed a RMSE of up to 0.35 mm. Additionally, this method enables the acquisition of stiffness information of the target, unlike using the prior image information. Therefore, it is expected that a stiffness overlay can be constructed and used for the diagnosis and treatment of targets.
