Comparison of the Image-to-Patient Registration Performance of a Novel Augmented Reality–Based and a State-of-the-Art Infrared Optical Neuronavigation System

BACKGROUND AND OBJECTIVES: Augmented reality-based neuronavigation systems (AR-NS) have been proposed as a potential addition to regular neuronavigation systems, such as optical or electromagnetic systems. Accurate image-to-patient registration is a prerequisite for clinical acceptance and implementation of AR-NS in neurosurgical workflows. In this study, the registration performance of a novel AR-NS using a head-mounted display with an integrated red-green-blue camera was compared with that of a state-of-the-art infrared optical neuronavigation system. The performance was evaluated in controlled phantom experiments and in a clinical setting, focusing on registration error, localization precision, and registration time. METHODS: Phantom experiments were performed using a human head model equipped with 8 fiducial markers under controlled conditions. Subsequently, adult patients who were undergoing cranial neurosurgery with preoperative T1-weighted contrast-enhanced MRI were included in the clinical study. In both phantom and clinical settings, the target registration error (TRE) and fiducial registration error were assessed using 7 registration fiducials and one independent target fiducial. The registration time was also recorded. Paired t-tests were used for the statistical analysis. RESULTS: Data from 37 patients were included for analysis. In phantom experiments, sub-3-mm TRE and sub-1-mm point localization were achieved with both systems (P = .63). In the clinical setting, mean TRE values for AR-NS and infrared optical neuronavigation system were 5.5 ± 3.0 mm [95% CI: 4.8, 6.2] and 4.0 ± 2.0 mm [95% CI: 3.4, 4.7], respectively. The mean TRE difference was 1.2 ± 3.6 mm [95% CI: 0.1, 2.4], P = .047. CONCLUSION: Comparable registration performance was observed in the phantom measurements. A statistically significant difference in the registration performance of the 2 systems was observed in the patient measurements. The difference between the phantom and clinical findings highlights that similar registration performance can be achieved by the 2 systems under controlled circumstances, but the effect of clinical variables on the registration performance requires further investigation.