gms | German Medical Science

ESBS 2005: Skull Base Surgery: An Interdisciplinary Challenge
7th Congress of the European Skull Base Society held in association with
the 13th Congress of the German Society of Skull Base Surgery

18. - 21.05.2005, Fulda, Germany

Accuracy of neuronavigation for the transsphenoidal approach to the sellar region

Meeting Contribution

Search Medline for

  • B.M. Hoelper - Department of Neurosurgery, Klinikum Fulda, Academic Hospital of the Philipps University Marburg, Fulda, Germany
  • R. Behr - Department of Neurosurgery, Klinikum Fulda, Academic Hospital of the Philipps University Marburg, Fulda, Germany

ESBS 2005: Skull Base Surgery: An Interdisciplinary Challenge. 7th Congress of the European Skull Base Society held in association with the 13th Congress of the German Society of Skull Base Surgery. Fulda, 18.-21.05.2005. Düsseldorf: German Medical Science GMS Publishing House; 2009. Doc05esbs52

DOI: 10.3205/05esbs52, URN: urn:nbn:de:0183-05esbs521

Published: January 27, 2009

© 2009 Hoelper et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( You are free: to Share – to copy, distribute and transmit the work, provided the original author and source are credited.




A high correctness of neuronavigation (NN) systems is essential for a save routine clinical application in surgery of the skull base. This mostly depends on the geometrical quality of the acquired data set (CT, MRI), registration process (marker, landmarks, surface) and technical accuracy. The mean topographic accuracy of current NN systems for cranial application is described in the range of 0,3–4 mm, but there is a lack of validating NN to fluoroscopy. Little is known about the accuracy of neuronavigation compared to fluoroscopy in cranial applications and despite all attempts to minimize sources of inaccuracies, each individual operative procedure owns the risk to be accompanied by an inaccurate neuronavigation procedure despite high technical accuracy. Therefore, this study evaluates the accuracy of a passive marker based frameless neuronavigation system in the sellar region by comparing definite anatomical landmarks in fluoroscopy and neuronavigation using a special designed fusion algorithm for MRI image fusion.


Twelve patients (6 female, 6 male, mean age 57 years) with mass lesions in the intra- and suprasellar region were prospectively included into this study. A transsphenoidal approach was indicated for tumour resection or biopsy in all cases. Six markers were fixed bifrontally to the head in a semicircular fashion. Thereafter, in all patients two 3D MRI data sets were acquired using a 1,5 Tesla MRI scanner (Gyroscan ACS-NT, Philips Medical Systems, Eindhoven, the Netherlands): a T1 gradient echo (TE=4,6, TR=20, +Gd) and a T2 turbo spin-echo (TE=120, TR=2416) weighted MRI data set. 6 of 12 patients received an additional a 2 mm slice spiral CT scan (Siemens Somatom Plus, Siemens AG, Erlangen, Germany). The image data were transferred to the neuronavigation system (BrainLAB Vector Vision 2, BrainLAB AG, Heimstetten, Germany) via optical disc. Both MRI data sets and the CT-scan were fused using the VectorVision2 Volume of Interest image fusion software (Image-Fusion Version 1.18ß137, BrainLAB AG, Heimstetten, Germany). After the patient’s head was fixed in a three pin head holder, the markers were registered. The fluoroscopic C-arch (Philips, GE Series 9400) was brought into line to the anterior skull base in sagittal view. After operative exposure of the sphenoid sinus, a pointer tracked by neuronavigation marked the anterior and posterior border of the sinus. From each pointer position a screenshot of the neuronavigation and simultaneously a fluoroscopic printout was taken. In both images the sphenoid sinus as well as the top of the pointer were identified. Then, the distance between the screenshot from the neuronavigation and the pointer position in the printout from the fluoroscopy was measured.


The preoperative mean registration accuracy is 1,8 mm and for all cases <2,5 mm. In two patients, the posterior border of the sphenoid sinus could not exactly be identified due to overlapping of adjacent bony skull base in fluoroscopy. Thus, distance between neuronavigation and fluoroscopy is be measured at the posterior border of the sphenoid sinus in nine patients, while this is possible in all twelve cases at the anterior border of the sphenoid wing. For the anterior rim of the sphenoid sinus (n=12), the mean deviation between neuronavigation and fluoroscopy in anterior-posterior (ap) direction is 2,91±1,77 mm (maximum 5,85 mm), in medial-lateral direction 1,83±1,11 mm (maximum 4,05 mm), and the mean euclidic distance (ED) is 3,45 mm±1,76 mm (Figure 1 [Fig. 1]). The maximal euclidic distance between both imaging methods is 7,12 mm. At the posterior rim of the sphenoid sinus (n=9), the mean deviation of this anatomical landmark between both imaging methods in anterior-posterior direction is 2,35±2,20 mm (maximum 7,88 mm) and 2,38±1,62 mm (maximum 5,85 mm) in medial-lateral direction. Euclidic distance is in mean 3,60±2,33 mm and maximal 8,08 mm. No significant difference was found between the anterior and posterior border of the sphenoid sinus. The preoperative registration accuracy did not correlate to the difference of the sphenoid sinus landmark marked in the fluoroscopy and neuronavigation. To analyze the image fusion accuracy, 975 landmarks were identified for both T1- and T2-weighted images. A significant difference in registration accuracy (p<0,01) for all landmarks between a complete volume fusion (CV) 1,6±1,2 mm and a volume of interest fusion (VOI) 0,7±1,0 mm was found. Neural network analysis predicted a higher accuracy of VOI image fusion (0,05–0,15 mm) compared to CV fusion (0,9–1,6 mm)


An approach to the sphenoid sinus can be performed safely guided by NN without the assistance of fluoroscopy. This helps to reduce x-ray exposure, but a mean inaccuracy of 3,45–3,60 mm with maximum of 8,1 mm in the sphenoid sinus implies not to rely on NN but to carefully identify anatomical and pathological structures. Furthermore, an excellent registration accuracy does not necessarily mean a good intraoperative landmark correspondence between fluoroscopy and NN. Inaccuracies of NN is mostly based on variations in marker based registration and image set accuracy, while technical errors may only play a minor role. Inaccuracies in MRI data sets (seen between air and tissue seen in MRI gradient echo sequences) with high magnetic susceptibility changes can be decreased by the volume of interest fusion which excludes those areas with high susceptibility changes.