A real-time metric for quantifying registration stability in mixed reality neurosurgical navigation

Abstract Objective Mixed reality (MR) guidance enables real-time superimposition of virtual anatomical models onto the surgical field but is inherently affected by spatial instability due to tracking limitations, sensor noise and head-mounted display dynamics. These frame-to-frame fluctuations may lead to transient mislocalisation of neuroanatomical structures and are not adequately captured by conventional static registration metrics. This study aimed to develop and validate a clinically interpretable metric for continuous assessment of registration stability in MR-guided neurosurgery. Methods The Registration Quality Index (RQI) was introduced as an automated, surface-based metric quantifying virtual–physical alignment through pixel-level boundary proximity analysis. A higher RQI value indicates greater boundary displacement, corresponding to worse registration quality. Validation was performed using a rigid skull phantom, a Varjo XR-3 mixed reality headset and ArUco-based optical tracking across fifteen experimental trials and seventy-five viewing configurations. RQI values were compared with established neuronavigation accuracy metrics, including Fiducial Registration Error (FRE) and Target Registration Error (TRE). Results Strong correlations were observed between RQI and FRE ( r = 0.89 (95% CI 0.69–0.96), p < 0.001) and between RQI and TRE ( r = 0.93 (95% CI 0.80–0.98), p < 0.001). Regression analysis demonstrated that each 1% increase in RQI corresponded to a 0.195 mm increase in FRE and a 0.244 mm increase in TRE. Real-time computation at 30 Hz was achieved without workflow disruption. RQI variability was significantly influenced by viewing angle, with lateral perspectives yielding higher displacement values than frontal views. Conclusions As a technical proof of concept, RQI enables continuous, automated monitoring of registration stability in MR-guided neurosurgical navigation, addressing a critical limitation of current static accuracy assessments. While the current proof-of-concept validation on rigid phantoms demonstrates strong construct validity, the interpretability ranges reported here are exploratory and further clinical studies involving in vivo conditions are required to establish clinically actionable decision boundaries and assess performance in the presence of brain shift and soft tissue deformation.