1Division of Neurosurgery, Department of Clinical Neurosciences, Geneva University Hospitals; and.
2Faculty of Medicine, University of Geneva, Switzerland.
Neurosurg Focus. 2021 Jan;50(1):E17. doi: 10.3171/2020.10.FOCUS20789.
Augmented reality (AR) in cranial surgery allows direct projection of preregistered overlaid images in real time on the microscope surgical field. In this study, the authors aimed to compare the precision of AR-assisted navigation and standard pointer-based neuronavigation (NV) by using a 3D-printed skull in surgical conditions.
A commercial standardized 3D-printed skull was scanned, fused, and referenced with an MR image and a CT scan of a patient with a 2 × 2-mm right frontal sinus defect. The defect was identified, registered, and integrated into NV. The target was physically marked on the 3D-printed skull replicating the right frontal sinus defect. Twenty-six subjects participated, 25 of whom had no prior NV or AR experience and 1 with little AR experience. The subjects were briefly trained in how to use NV, AR, and AR recalibration tools. Participants were asked to do the following: 1) "target the center of the defect in the 3D-printed skull with a navigation pointer, assisted only by NV orientation," and 2) "use the surgical microscope and AR to focus on the center of the projected object" under conventional surgical conditions. For the AR task, the number of recalibrations was recorded. Confidence regarding NV and AR precision were assessed prior to and after the experiment by using a 9-level Likert scale.
The median distance to target was statistically lower for AR than for NV (1 mm [Q1: 1 mm, Q3: 2 mm] vs 3 mm [Q1: 2 mm, Q3: 4 mm] [p < 0.001]). In the AR task, the median number of recalibrations was 4 (Q1: 4, Q3: 4.75). The number of recalibrations was significantly correlated with the precision (Spearman rho: -0.71, p < 0.05). The trust assessment after performing the experiment scored a median of 8 for AR and 5.5 for NV (p < 0.01).
This study shows for the first time the superiority of AR over NV in terms of precision. AR is easy to use. The number of recalibrations performed using reference structures increases the precision of the navigation. The confidence regarding precision increases with experience.
增强现实（AR）技术可在手术显微镜的实时视场中直接叠加显示预先注册的图像。本研究旨在通过模拟手术条件下的颅骨，比较 AR 辅助导航和标准指针式神经导航（NV）的精确性。
对商业化的标准 3D 打印颅骨进行扫描、融合，并与患者的磁共振（MR）图像和 CT 扫描进行参考融合，该患者右侧额窦存在 2×2mm 的缺损。我们对缺损进行了识别、注册，并整合到 NV 中。在 3D 打印颅骨上物理标记出与右侧额窦缺损相匹配的目标。共有 26 名参与者参与了研究，其中 25 名没有 NV 或 AR 的使用经验，1 名仅有少许 AR 使用经验。参与者接受了 NV、AR 和 AR 重新校准工具的简要培训。参与者被要求完成以下任务：1）“在 3D 打印颅骨上仅使用 NV 定向导航指针定位缺损中心”，2）“在常规手术条件下使用手术显微镜和 AR 聚焦于投影目标的中心”。对于 AR 任务，记录了重新校准的次数。参与者在实验前后使用 9 级 Likert 量表评估了对 NV 和 AR 精度的信心。
AR 的目标距离中位数明显低于 NV（1mm[Q1：1mm，Q3：2mm]比 3mm[Q1：2mm，Q3：4mm]，p＜0.001）。在 AR 任务中，中位数的重新校准次数为 4 次（Q1：4 次，Q3：4.75 次）。重新校准的次数与精度显著相关（Spearman rho：-0.71，p＜0.05）。实验后的信任评估中，AR 中位数为 8，NV 中位数为 5.5（p＜0.01）。
本研究首次表明 AR 在精度方面优于 NV。AR 易于使用。使用参考结构进行的重新校准次数越多，导航精度越高。经验越丰富，对精度的信心就越高。
