gms | German Medical Science

Objective: Mapping image space and physical space in neuronavigation is most commonly performed using fiducial based imaging approaches based on MR or CT images, whereas CT images show superior registration accuracy due to image distortion within the MR images. Fiducial based registration for navigation has an overall clinical application accuracy of 3-4 mm. The aim of this study was to compare fiducial based and automatic registration using the AIRO iCT scanner (Brainlab, Munich) regarding registration accuracy analyzing a wide range of CT protocols.

Methods: To investigate registration accuracy with different scan protocols, a plexiglas phantom was used consisting of a base plate and 24 randomly distributed notched rods on varying length. To perform both registration approaches, also 7 additional adhesive fiducial markers were attached diffusely scattered. A reference data set (helical, full dose scan) was acquired, to allow for defining the target points of each notched rod in the image space. Fiducial based registration was performed for axial and helical CT data (160 mA, 120 kV). Automatic registration was performed using scan protocols with varying amperage (160 mA, 80 mA, 40 mA, 20 mA, 10 mA, 5 mA) and voltage (120 kV, 100 kV, 80 kV) and all notched rods were localized and acquired with the tool tip of a pointer. For each notched rod the Euclidean distance between reference point and acquired point was calculated as target registration error (TRE).

Results: Standard fiducial based registration showed an overall TRE of 1.15 ± 0.43 mm. Automated axial iCT based registration revealed an overall TRE of 1.08 ± 0.18 mm ranging from 0.86 mm to 1.24 mm, showing comparable (p > 0.05) or significantly improved mean TRE results (p < 0.05) depending on the scan protocol. Automated helical iCT based registration revealed an overall mean TRE of 0.47 ± 0.23 mm, ranging from 0.34 mm to 0.70 mm, showing a significantly improved (p < 0.05) mean TRE for all scan protocols compared to fiducial based registration.

Conclusion: ICT-based automatic registration delivers even in this optimal set-up for fiducial based registration (e.g. no soft tissue shift during data acquisition and registration) in case of axial scans at least comparable or even significantly better registration accuracy, in case of helical acquisition, significantly better registration accuracy is achieved across all scan protocols and is therefore a useful tool for user-independent and fast, highly accurate registration in neurosurgery.