Restricted surface matching a new approach to registration in computer assisted surgery

Abstract

Since the first introduction of tumor stereotaxis to neurological surgery in the early 1980's, computer assisted surgery (CAS) has been applied to several medical fields. An important common issue for all CAS systems is the registration between preoperative data (e.g. computed tomography (CT), magnetic resonance (MR) images, ...) and intraoperative data. In our application the objective is to obtain position information of medical tools with respect to CT images of bony structures. We describe in this paper a new approach for the registration of preoperative CT images with the patient in the operating room. The method, that we named ‘restricted surface matching’, is a fast, accurate, and robust surface matching based method.

Preoperatively, a three-dimensional (3D) distance map is created from the segmented images and 3–5 coarse anatomical landmarks are defined on the relevant surface. Intraoperatively, the surgeon uses an optically tracked pointer to first digitize the 3D coordinates on the bone corresponding to the predefined coarse landmarks then another 10–15 random points on the surface of the bone. A matching is then performed between the real world and image data using a minimization algorithm. The coarse landmarks are used as restricting parameters to allow a fast convergence towards the global minimum.

There are two major improvements of our approach compared to other surface matching based methods: (a) only a few points have to be captured by the surgeon and (b) the coarse landmarks do not have to be captured exactly, so deviations from the precise location do not matter as it might be difficult to find (e.g. on the pelvic bone).

A perturbation analysis showed that the program was stable and not sensitive to noise. The accuracy of the method was evaluated on a cadaveric bone specimen to be 0.3 mm average over ten trials. The algorithm was used with success in the operating theater for three periacetabular osteotomy (PAO) surgeries.