STUDY DESIGN In this cadaveric study, a computer-assisted image guidance system was tested for accuracy of thoracic pedicle screw placement. OBJECTIVES Evaluate the system's accuracy for thoracic pedicle screw placement in vitro. SUMMARY OF BACKGROUND DATA The effective use and reliability of pedicle screw instrumentation in providing short-segment stabilization and correction of deformity is well known in the lumbar spine. Pedicle screw placement in the thoracic spine is difficult because of the small dimensions of the thoracic pedicles and risk to the adjacent spinal cord and neurovascular structures. Investigators have shown the improved accuracy of computer-assisted lumbar pedicle screw placement; but the accuracy of computer-assisted thoracic pedicle screw placement, which is becoming more widely used, has not been shown. METHODS In five human cadavers, 120 thoracic pedicle screws were placed with computer-assisted image guidance. The largest clinically feasible screw was used based on the cross-sectional dimensions of each pedicle. The accuracy was assessed by postoperative computed tomography and visual inspection. RESULTS The overall pedicle cortex violation was 23 of 120 pedicles (19.2%). Nine violations (7.5%) were graded as major and 14 (11.7%) as minor. A marked and progressive learning curve was evident with the perforation rates that decreased from 37.5% in the first cadaver to 4.2% in the last two cadavers. CONCLUSIONS Accurate thoracic pedicle screw placement is feasible with computer-assisted surgery. However, as with any other new surgical technology, the learning curve must be recognized and incorporated into the necessary fundamental knowledge and experience for these procedures.