Markerless registration for intracerebral hemorrhage surgical system using weighted Iterative Closest Point (ICP)

It is required to use a stereotactic frame on a patient's crainial surface to access an intracerebral hematoma in conventional ICH (Intracerebral Hemorrhage) removal surgery. Since ICH using a stereotactic frame is an invasive procedure and also takes a long time, we attempt to develop a robotic ICH removal procedure with a markerless registration system using an optical 3-D scanner. Preoperative planning is performed using a patient's CT (Computed Tomography) images, which include the patient's 3-D geometrical information on the hematoma and internal structures of brain. To register the preplanned data and the intraoperative patient's data, the patient's facial surface is scanned by an optical 3-D scanner on the bed in the operating room. The intraoperatively scanned facial surface is registered to the pose of the patient's preoperative facial surface. The conventional ICP (Iterative Closest Point) algorithm can be used for the registration. In this paper, we propose a weighted ICP in order to improve the accuracy of the registration results. We investigated facial regions that can be used as anatomical landmarks. The facial regions for the landmarks in the preoperative 3-D model are weighted for more accurate registration. We increase weights at the relatively undeformed facial regions, and decrease weights at the other regions. As a result, more accurate and robust registration can be achieved from the preoperative data even with local facial shape changes.