Sci—Thur PM: Imaging — 02: Repeated landmark use for patient‐to‐image registration reduces fiducial registration error in patient‐to‐image mapping in image guided neurosurgery

The patient‐to‐image mapping for an image guided neurosurgical procedure is traditionally determined through the identification of 9 anatomical landmark pairs on both diagnostic preoperative magnetic resonance images and the patient in the operating room. In this study we investigate the effect of using an increased number of point pairs on the mean fiducial registration error of the landmark registration. This was first evaluated in the lab on a custom built precision milled linear testing apparatus. It was used in conjunction with an optical tracking system and a tracked surgical pointer. A volume of 700 points was registered and showed a plateau of the mean fiducial registration error when a large number of points were used. This was then extended to the operating room where 5 sets of the 9 anatomical landmarks were used for registration and compared to the registration only done with 1 set. A significant improvement (one tailed t‐test) was seen in five of the six cases showing an improvement in patient‐to‐image registration with no significant addition to operation duration.