Clinical Validation of Intraoperative Ultrasound-Based Automated Rigid Image Fusion to Update Neuronavigation Using Preoperative MRI.

Abstract

Background and objectives: MRI-based neuronavigation may suffer from inaccuracies that can be compensated by navigated 3D intraoperative ultrasound (iUS) and applying MRI-iUS rigid image fusion (RIF). In this work, such an automated application is evaluated.

Methods: Twenty-five adult patients with gliomas were enrolled and underwent resection using navigated iUS. Intraoperative evaluation and postoperative quantification [ie, measurement of the target registration error (TRE)] were conducted to assess the accuracy of registration-based fusion and automated RIF at various stages of surgery [before dura opening (BDO), after dura opening (ADO), after partial resection (APR), after completion of resection (ACR)]. Linear mixed models were used to assess and analyze TRE and the effect of patient- and tumor-related factors on the performance of the RIF. Furthermore, the TRE was measured after applying different prealignments.

Results: In total, 79 MRI-iUS data sets derived from 24 patients and enriched with 600 anatomic landmark pairs were evaluated. Overall, RIF resulted in a significantly reduced mean TRE compared with registration-based fusion (from 4.7 mm to 3.5 mm, P < .002). This difference in TRE was dependent on the stage of surgery, being significant for BDO, ADO, and APR stages, but not ACR. It was independent of any tumor-related factors. Simulation tests showed that RIF can significantly improve TRE for a range of ±15 mm prealignment accuracy with highest effect for BDO and ADO.

Conclusions: RIF using intraoperative navigated ultrasound improves registration accuracy for intra-axial tumor surgeries. It shows reliable results not only for preresection stages but also partially for later surgical stages.