Evaluation of registration methods used in frameless stereotactic surgery for the lumbar and cervical regions of the spine.

Computer-assisted pedicle screw insertion is feasible but has proved to be problematic. The purpose of this study was to detail the accuracy of registration techniques and pedicle screw insertion using a frameless stereotactic system. Two registration techniques were evaluated on a model spine. The frameless stereotactic system was then used to insert 26 pedicle and 8 lateral mass screws in human cadavers. For posterior vertebral elements, trajectory accuracy was 2.5 +/- 1.0 mm between T12 and L5 and 2.2 +/- 0.9 mm between C2 and T1. Registration of the anterior elements, however, was less accurate. Despite this flaw, all screws were inserted without penetrating the cortex. Screw trajectory was accurate to 2 degrees. The main limitation of frameless stereotactic surgery in the spine stems from the fact that only the posterior vertebral elements are used during registration. Despite this flaw, the system placed all screws correctly. Given these limitations, we believe that this system is most useful for locating the screw insertion point and providing a trajectory in the pedicle.