P274. Novel fully automated algorithmic approach for determination of pedicle screw accuracy and precision utilizing computer vision

BACKGROUND CONTEXT Historically, pedicle screw accuracy measurements have relied on computed tomography (CT) and expert visual assessment of the position of pedicle screws relative to preoperative plans. Proper pedicle screw placement is necessary to avoid additional complications and cost of revision procedures. PURPOSE To determine accuracy and precision of pedicle screw insertion via a novel computer vision algorithm utilizing preoperative and postoperative CT scans. STUDY DESIGN/SETTING Cadaveric controlled lab study. PATIENT SAMPLE Three adult cadaveric specimens. OUTCOME MEASURES The difference in tip, tail, and mid-pedicle screw distance (in millimeters), mean error, variance, and mean absolute error between preoperative plan and postoperative CT. METHODS Screw placement was planned in generic planning and segmentation software using standard criteria for screw planning. Two surgeons experienced in robotic workflow performed bilateral T2-L4 instrumentation using robotic-assisted navigation. Postoperative CT scans of the entire vertebral column were obtained. Automated segmentation and computer vision techniques were employed to align each postoperative vertebra with its preoperative counterpart and then compare positions along all three axes in an automated fashion. The mid-pedicle position in the anterior-posterior plane was not reported as it is not a measurable parameter. Mean error (ME), mean absolute error (MAE), and variance were calculated. Systematic error, precision, and accuracy were defined as ME, 2 standard deviations, and MAE, respectively. RESULTS Eighty-eight pedicle screws were placed. The computer algorithm showed that robotic-assisted pedicle screw placement was both accurate and precise. This study demonstrates robotic assisted pedicle screws can be placed with submillimeter accuracy in the coronal plane. Furthermore, the tip measurements showed greater precision than the tail measurements. Figure 1 shows both the numerical values and a visual representation of accuracy and precision. CONCLUSIONS This study introduces automated algorithms for determining accuracy and precision of planned pedicle screws. Our accuracy outcomes are comparable or superior to recent data from robotic-assisted in vivo studies in adult patients, as well as prior cadaver studies. This algorithmic computerized workflow establishes a standardized protocol for assessment of pedicle screw placement accuracy and precision, and provides baseline accuracy and precision for both cadaveric and in vivo comparison. FDA Device/Drug Status Medtronic Mazor X Stealth (Approved for this indication)