Navigation system

A frameless navigation system (ACCISS II™, Schaerer Mayfield Technologies GmbH, Berlin, Germany) was used for intraoperative image guidance in all cases. The system comprises the hard- and software necessary to generate and detect a DC pulsed magnetic field for computing the position and orientation of a localizing sensor. The tracking system in its basic version consists of an electromagnetic transmitter unit, a sensor (which is integrated into the handle of a surgical pointer device) and an electronic digitizer unit that controls the transmitter and receives the spatial data from the localizing sensor.

The transmitter consists of a triad of electromagnetic coils (size: 9.6 cm cube) which generates a homogeneous electromagnetic field (max. 600 milligauss with a translation range of 76.2 cm in any direction) that, in its basic version, simultaneously serves as the fixed reference for the setup.

The localizing sensors can be integrated into pointers or other surgical instruments of various shapes. The sensor, being completely passive and having no active voltage applied, detects the magnetic field generated by the transmitter unit with up to 120 measurements per second what ensures real-time conditions. They have 6 degrees of freedom (position and orientation) with an angular range of ± 180° azimuth & roll and ± 90° elevation. The static accuracy is specified by the manufacturer (Ascension Technologies Corp., Burlington, USA) as 1.8 mm RMS (position) and 0.5° RMS (orientation). The static resolution is 0.5 mm (position) and 0.1° (orientation) at a distance of 30.5 cm from the transmitter.

In the digitizer unit, the analogue measured signals of the sensor are digitalized, and the coordinates of the sensor position are calculated.

Dynamic Reference Frame (DRF)

To allow simultaneous registration, localization and position tracking of more than one localizing sensor, the before described basic version of the ACCISS II system was expanded with a soft- and hardware update which helps to run a so-called Dynamic Reference Frame (DRF). The DRF can be used as an additional reference system that defines an independent coordinate system in space in addition to the one established by the transmitter unit (Figure 1B). Thus, it becomes possible to record the slightest movement of the cranium as well. This information can then be used to continuously adapt the position of the imaging plane and the resultant calculated virtual 3-D model to the actual position of the cranium. Technically, the DRF consists of an additional localizing sensor measuring 8 mm × 8 mm × 18 mm in size with a weight of 1.2 g. The extra sensor is accommodated in a watertight capsule and is connected to the navigation system with a 3 m long cable. The DRF sensor can either (a) be attached to a dental splint or (b) be attached retroauricular on the hairless skin.

1. (a)

   In the oral cavity, the DRF is attached to the upper row of teeth using a special, removable mouthpiece (Figure 2) and a 2-component polyether self-hardening material (Impregum^® F; ESPE Dental AG, Seefeld, Germany). The mouthpiece consists of a U-shaped splint which is filled with a fast-hardening material and applied to the upper row of teeth exerting slight pressure (about 0.25 atm) at the centre. The vacuum resulting from hardening of the material ensures that the mouthpiece is firmly secured in place in patients with healthy teeth. After the procedure, the mouthpiece is removed by releasing the vacuum with a dental hook.

    

Proper affixation of the DRF was checked in all cases by a rotation test immediately after image data registration (Figure 3). To this end, the head was rotated about 120° from the right lateral to the left lateral position and back (Figure 3A–C). The spatial coordinates of the fiducial markers were verified relative to the position of the DRF. Adequate attachment of the DRF was assumed when the deviation was < 1 mm in all three spatial direction (cartesian x, y, z-coordinates as displayed by the navigation system; Figure 3D). If there was greater deviation, the position was corrected and the attachment optimized until deviation was within the limit of 1 mm.

Image data acquisition and preparation

Preoperatively, a serial CT or MRI scan was obtained. The images consisted of a three-dimensional volume data set of contiguous axial CT or sagittal MR images. In order to obtain isotropic voxels of 1 mm length one of the following CT or MRI protocols was routinely used.

MRI was performed using a T1-weighted 3D GE sequences (3D MP RAGE) with the parameters: TR 9.7 ms, TE 4 ms, FA 12°, TI 300 ms, TD 0 s, FOV 256 mm, 256 × 256 matrix, 256 partitions, slice thickness of 1 mm, acquisition time 11 min 54 s. Alternatively, a high-resolution CT spiral scan was acquired with 1 mm slice thickness, 512 × 512 matrix, pitch factor 2, 1 mm increment, and 50–110 mA tube current. The image data were transferred to the computer workstation in the ACR/NEMA 3.0/DICOM image data format via a local network (LAN – FTP or DICOM transfer protocol), or through data media, such as CD-ROM, magnetic-optical disks (MOD) or magnetic tape (DAT).

Data processing and preparation was performed using an autosegmentation technique (ACCISS II software version 1.9). Image guidance was based on axial planar views (sagittal, coronal and transaxial), free planar views (defined by pointer orientation and/or target localization), and 3D views of the anatomical objects (skin, skull, brain surface structures, brain parenchyma and lesion target) (Figure 4). The image data was registered by means of point-to-point matching (sequentially sampling 7 two-component adhesive fiducial markers with a sensor-bearing pointer according to a standardized protocol).

Accuracy measurements

Registration accuracy was determined calculating the fiducial registration error (FRE) expressed as the root mean square error. The FRE describes the distance between the position of a marker in the image dataset and the position measured in the operative field. The mean RMS value is calculated directly by the navigation system and is displayed together with the min. and max. FRE and the Target Registration Error (TRE – for a certain target point within the registered volume) on the navigation screen (Figure 3D).

The application accuracy was monitored intraoperatively using as a reference point a 1 mm burr hole drilled into the exposed bone margin directly after craniotomy (Figure 5). The initial Cartesian coordinates of this reference point were determined immediately by means of a pointer. The measurements were repeated after craniotomy immediately before dura opening, three times during tumour resection (M1–M3) and after closure of the dura, respectively at the end of the operation. Deviations in x, y, and z directions were measured as three-dimensional Position Error (PE in mm) of the reference point relative to the baseline coordinates of the same point determined immediately after craniotomy.

Statistical analysis

FRE and PE values are expressed as means +/- standard deviation from the number (n) of patients in each group. Data were tested for significance using one-way ANOVA to determine degree of variability within a group, followed by Bonferroni post hoc analysis. Test of pairwise comparisons were carried out with the Student's t-test to compare two groups (e.g. for differences in FRE between the different types of head fixation, as well as for differences in ΔPE in-/decrease between the different types of head positioning over the time of surgery). A p < 0.05 was considered as statistically significant. Data management and statistical analyses were performed using the SPSS 13.0 for Windows^® software package.

Patients and indications for frameless stereotaxy

The clinical study included 40 patients with intracerebral tumours or lesions in the area of the skull base in whom intraoperative navigation was used to localize the target or trajectory or to determine the extent of resection. The patients had the following diagnoses: 2 WHO II gliomas, 8 WHO III gliomas, 7 glioblastomas, 12 metastases, 2 primary bone tumours, 2 meningiomas, 4 lymphomas, 2 fibrous dysplasia, and one paraganglioma. There were 22 men and 18 women with a mean age of 55.7 years (range 18 – 81 years). CT data sets were used for navigation in 9 cases and MRI data sets in the remaining 31 patients. All steps of the examinations were approved by the institutional review board. Written informed consent was available from all patients participating in the study. The interventions were performed at the Department of Neurosurgery, Charité – Universitätsmedizin Berlin, Campus Benjamin Franklin.

Clinical application

According to the indication for the use of intraoperative navigation, the patients were assigned to one of three types of interventions according to head fixation and use of the DRF including its mode of attachment (Figure 6). To assign the patients to one of the three groups, the following questions were answered: „Is it planned to reposition the patient/the patient's head during the operation?” and „Is it likely that there will be involuntary head movement during certain surgical manoeuvres?”

Type A comprised 10 patients in whom no intervention-related repositioning was planned and in whom involuntary movement of the head was unlikely because 3-point pin fixation was used. These patients were operated on with navigation performed under standard conditions and without additional use of the DRF. The patients of this group served as controls.

Type B consisted of two subgroups. The first subgroup included those cases in whom intraoperative repositioning of the head was planned. These were 6 patients scheduled for removal of two lesions in one session. All 6 patients were actually repositioned during the operation. The other 4 patients assigned to this group had large lesions at the skull base, making it likely that voluntary or involuntary changes in head position would occur during the intervention. All 10 patients of this group were operated on using 3-point pin fixation in combination with a DRF. The DRF was attached orally (type B1) in 5 cases (Figure 7) and retroauricularly in the other 5 cases (type B2).

Type C consisted of those cases in whom intraoperative head movement was expected or desirable as well as those patients in whom repositioning might have become necessary in the course of the operation. These were 13 patients scheduled for burr hole procedures for neuroendoscopic interventions or biopsy, 3 patients in whom a transnasal approach was planned, and 4 patients undergoing awake craniotomy for removal of lesions from language areas. In 10 patients of this group, the DRF was attached in the oral cavity (type C1); in the other 10 cases, retroauricular attachment was necessary because of the dental status or for anesthesiological reasons and in the patients who underwent awake craniotomy to perform intraoperative speech testing (type C2).

Registration accuracy

The mean fiducial registration errors (FREs) were 1.51 mm (+/- 0.36 mm SD) in the control group type A (n = 10), 1.56 mm (+/- 0.40 mm SD) in type B1 interventions (n = 5), 1.54 mm (+/- 0.33 mm SD) in type B2 (n = 5), 1.73 mm (+/- 0.63 mm SD) in type C1 (n = 10), and 1.75 mm (+/- 0.41 mm SD) in type C2 (n = 10) (Figure 8). There was no statistically significant difference between rigid pin fixation (r.p.f.) of the head alone (control group: type A) and r.p.f. with additional DRF (type A vs. type B1; p > 0.05 and type A vs. type C2, p > 0.05). In DRF-supported procedures, there was no statistically significant difference between oral and retroauricular placement of the DRF, neither in cases with rigid pin fixation (type B1 vs. type B2, p > 0.05), nor in the unfixed head mode (type C1 vs. type C2, p > 0.05). However, both types of unfixed head positioning (type C1 and C2) presented with significant higher FRE mean values compared to control type A (type A vs. type C1; p < 0.05 and type A vs. type C2; p < 0.05), as well compared to both groups of r.p.f. with additional DRF (type C1 vs. B1, p < 0.05; type C1 vs. B2, p < 0.05; type C2 vs. type B1, p < 0.05 and type C2 vs. type B2, p < 0.05).

Application accuracy

The mean position errors (PEs) measured after completion of craniotomy and before dura opening (on average 71 min after the end of image data registration) were 0.79 mm (+/- 0.23 mm SD) in type A interventions, 0.71 mm (+/- 0.18 mm SD) in type B1, 0.93 mm (+/- 0.34 mm SD) in type B2, 0.98 mm (+/- 0.31 mm SD) in type C1, and 0.69 mm (+/- 0.25 mm SD) in type C2. The mean PEs measured at 3 consecutive time points during tumour resection (M1= about 20 min after dura opening/M2= about 45 min after dura opening/M3= about 70 min after dura opening) were 0.95/1.28/1.34 mm in type A, 0.95/1.17/1.20 mm in type B1, 1.10/1.31/1.44 mm in type B2, 1.19/1.58/1.74 mm in type C1, and 1.03/1.44/1.64 mm in type C2. The final measurements for determining application accuracy at the time of dura closure or at the end of the intervention (on average 84 min after dura opening) yielded mean PEs of 1.45 mm (+/- 0.34 mm SD) for type A interventions, 1.26 mm (+/- 0.29 mm SD) for type B1, 1.44 mm (+/- 0.30 mm SD) for type B2, 1.86 mm (+/- 0.29 mm SD) for type C1, and 1.68 mm (+/- 0.38 mm SD) for type C2 (Figure 9). ΔPEs (measured as difference in mm between the initial PE at the time of dura opening and the corresponding PE at the time of dura closure) were significantly higher in types C1 and C2 compared with control group (types A) (* p < 0,05, t test). There were no statistical differences between the two B-type procedures and control group (type A) (p > 0.05).

Complications

Complications such as damage to the teeth or loosening resulting from oral attachment of the DRF were not observed in any of the 15 cases (types B1 and C1). Two patients showed small pressure sores on the lips due to direct contact with the plastic coating of the DRF. The sores healed spontaneously and without complications in the course of three days.

Retroauricular DRF attachment was also not associated with any major complications. None of the 15 patients in this group developed any pressure sores in the area of the auricle or the skin over the mastoid (types B2 and C2). The waterproof self-adhesive film prevented detachment of the tape, e.g. by disinfectant, in all 15 cases. One patient developed an allergic skin reaction to the tape used. The irritated skin responded well to topical ointment application and the irritation resolved within 24 hours.

Frameless stereotaxy without rigid-pin fixation of the head

Various techniques for identifying and compensating for intraoperative head movements during image-guided surgery have been published to date [12–18]. Ohhashi et al. [17], for example, reported on their experience with an electromagnetic navigation system originally developed for complex ENT operations of the sinus. This system comprises a plastic headset and does not require rigid head fixation. An advantage of this system described by its users is the fact that the headset simultaneously serves as a reference for automated image registration without the need for placement of fiducial markers. However, this system is geometrically restricted to the frontal region and the facial skull, ensuring an adequate navigational accuracy only in a limited volume around the headset, i.e. the area of the midface and the sinus. In contrast, the sensor-based DRF presented in this study enabled image data registration over the entire geometric area of the head where markers are placed. This is especially significant for intracranial operations as the number and size of the clusters registered directly affects the surgical accuracy in the respective area [9–11].

Another, really non-invasive attachment technique for a reference system, the so-called Vogele-Bale-Hohner mouthpiece, which moves with the patient's head was described by Bale and co-workers [18] in 2000. It acts as a reference arc that is attached to a tailored vacuum-affixed dental cast. The authors demonstrated that the accuracy achieved with their system is comparable to that of rigid pin fixation [18]. However, the size and shape of the reference frame that is also used for registration and the reliance on optical measurement also has drawbacks in that patient comfort is restricted and the system is of limited usefulness when the patient is positioned prone and the system thus comes to lie on the side away from the camera. Moreover, the device cannot be used in awake craniotomy with intraoperative language mapping since the mouthpiece does not allow the patient to speak. This problem can be overcome by the use of the sensor-based DRF and its positioning in the retroauricular area. Moreover, an electromagnetic system avoids the line-of-sight problem, i.e. it does not require an undisturbed visual contact between the DRF and a camera system. The here described electromagnetic technique thus allows unrestricted positioning as well as intraoperative repositioning of the patient. As the position of the sensor-bearing pointer is determined relative to the coordinate system defined by the DRF, the system operates in a virtual environment. Head movements and changes in position can be directly followed on the system's monitor in real-time. The same also holds true for image-guided instrument control. Here, very minute changes in position can directly affect the identification e.g. of a trajectory chosen on the basis of the image data.

Accuracy of frameless stereotaxy with the DRF in the clinical application

Fixation and attachment technique

The use of a dental splint attached to the upper row of teeth appears to be feasible in patients with healthy teeth. The self-hardening material used here is non-toxic, hypoallergic, fast-drying, and form-stable. Firm attachment is ensured as long as a vacuum is maintained between the hardened 2-component polyether system and the dental surface. Once the vacuum is released, e.g. with a small dental hook, the splint can be removed easily and rapidly. Others have described the ease of using such material intraoperatively. For example, Bale et al. [15, 16, 18] used a comparable polyether material for attachment of the mouthpiece of the Vogele-Bale-Hohner head holder. After two of our patients developed pressure sores on the lips, we have since placed a wet compress around the DRF following its placement. The compress protects the oral soft tissue and additionally prevents drying of the oral mucosa. No pressure sores have occurred thereafter.

Retroauricular attachment with adhesive tape was found to be inferior to oral attachment in terms of application accuracy, which is probably due to displacement of the DRF on the skin although no loosening of the tape was noted in any of the patients. Nevertheless, we think that a decrease in skin turgor and the weight of the sterile film are potential causes of the inaccuracies measured. With this potential source of inaccuracy in mind, one can still use retroauricular attachment of the DRF with benefit in patients with a poor dental status and in patients scheduled for awake craniotomy, as in patients undergoing awake craniotomy, pain associated with placement of the pins [20] and the risk of injuries through inadvertent head movements, less patient comfort, and declining ability of the awake patient to cooperate as the operation proceeds interfere with rigid fixation of the head.