Accuracy and reproducibility of automatic versus manual registration using a cone-beam CT image guidance system

Introduction: Intraoperative imaging reveals morphologic changes and resolves anatomic uncertainties during surgery. The automatic registration (AR) approach provides registered intraoperative images for real-time tracking within seconds of acquisition. Purpose: (1) To design an AR device for clinical use integrated with cone-beam computed tomography, (2) to compare the accuracy and reproducibility of manual and automatic registration, and (3) to evaluate the robustness of the AR system. Methods: An AR device consisting of an acrylic face shield with fiducials mounted on an adjustable arm was designed. Eight surface and five internal divot markers were placed with bony fixation to a cadaveric head. Internal markers were localized on the image representing the "true" location. This was compared to the positions localized using a navigational system when both manual registration and AR were applied. A series of surgical tasks and variation of the AR device height above the surgical field was performed, and target registration error (TRE) was measured. Results: The mean fiducial registration error (FRE) for manual and automatic registration was 0.72 mm ± 0.03 and 0.41 mm ± 0.01, respectively. The mean TRE for manual and automatic registration was 0.89 mm ± 0.26 and 0.91 mm ± 0.25, respectively. Conclusions: AR offers a more accurate and reproducible FRE and a TRE equally comparable to that of manual registration. This system also demonstrates robustness with comparable accuracy and reproducibility throughout different surgical tasks and variation of AR device height up to 9 cm above the surgical field. This system is currently being translated into clinical trials.