The aim of this study was to perform an objective assessment of the accuracy of mandibular osteotomy simulations performed using an image-guided sagittal saw. A total of 16 image-guided mandibular osteotomies were performed on four prefabricated anatomical models according to the virtual plan. Postoperative computed tomography (CT) image data were fused with the preoperative CT scan allowing an objective comparison of the results of the osteotomy executed with the virtual plan. For each operation, the following parameters were analyzed and compared independently twice by two observers: resected bone volume, osteotomy trajectory angle, and marginal point positions. The mean target registration error was 0.95 ± 0.19 mm. For all osteotomies performed, the mean difference between the planned and actual bone resection volumes was 8.55 ± 5.51%, the mean angular deviation between planned and actual osteotomy trajectory was 8.08 ± 5.50°, and the mean difference between the preoperative and the postoperative marginal point positions was 2.63 ± 1.27 mm. In conclusion, despite the initial stages of the research, encouraging results were obtained. The current limitations of the navigated saw are discussed, as well as the improvements in technology that should increase its predictability and efficiency, making it a reliable method for improving the surgical outcomes of maxillofacial operations.
Intraoperative navigation, also referred to as image-guided surgery (IGS) or computer-assisted navigation (CAN), continues to gain credibility in the field of craniomaxillofacial surgery. The main function of intraoperative navigation systems is to allow the surgeon to precisely locate the surgical instruments or bony anatomical landmarks in the three-dimensional (3D) surgical field in real-time. This feature, in conjunction with advanced preoperative virtual planning and options for the prediction of postoperative results, makes computer-assisted navigation technology a powerful tool facilitating surgical operations in the technically challenging head and neck region.
Despite the fact that in recent years there have been many reports describing the scope and range of current clinical applications of this technology in craniomaxillofacial surgery, as well as many accuracy tests on modern navigation system tip-pointers, there appears to have been no research to establish the level of accuracy of mandibular osteotomies performed using the image-guided surgical saw. Therefore, the purpose of this study was to perform an objective evaluation of the precision of mandible model osteotomies performed using the image-guided resection technique reported recently by the present authors.
Materials and methods
Image data acquisition
Thirteen titanium microscrews (diameter 1.0 mm, length 4.0 mm) were inserted into a plastic mandible model (type A20; 3B Scientific GmbH, Hamburg, Germany). The skull with the mandible model was scanned using a 32-slice CT scanner (Somatom Sensation 16; Siemens Medical Solutions, Erlangen, Germany) with the following parameters: the 512 × 512 pixels dataset was acquired at a resolution of 0.39 mm/pixel and 0.625 mm slice thickness. Image data were saved in Digital Imaging and Communication in Medicine (DICOM) format.
The DICOM data were transferred to a Windows-based computer workstation with the Maxillo-Facial Surgery System (MFSS), which was created by bioengineers and software engineers from Wroclaw University of Technology in cooperation with the Maria Sklodowska-Curie Memorial Cancer Centre and Institute of Oncology in Warsaw. Using the MFSS virtual planning module, these images were reformatted to produce standard two-dimensional (2D) axial, frontal, and sagittal views, as well as a 3D volume-rendered model. Using the bone segmentation option, four separate osteotomies of the mandible were planned ( Fig. 1 ). Next, the microscrew heads were identified manually and labelled as either registration fiducials for registration purposes or target fiducials for target registration error (TRE) estimation, individually for each virtual plan of the osteotomy. Each virtual surgical plan was saved and exported to the intraoperative navigation module of the MFSS system.
Printing of 3D models
Due to the fact that the mandible model was made of plastic and cutting it with a surgical saw would cause the osteotomy edges to melt, thus preventing a reliable analysis of the objective postoperative results, it was decided to produce an exact model reproduction made of plaster. Using the MFSS software, the mandible DICOM data were converted to stereolithography file (STL) format and sent to the CAD/CAM facility (Department of Laser Technology, Automation and Organization of Production, Wroclaw University of Technology, Wroclaw, Poland) where four exact plaster mandibles, with no significant loss of fiducial marker accuracy, were fabricated using generative 3D printing technology.
Image-guided model osteotomy
All surgical procedures were performed in a real operating theatre setting, according to the same operating protocol. A dynamic reference frame (StealthStation Spine Referencing Set; Medtronic, Minneapolis, MN, USA) was attached rigidly to the mandible and an optical tracking adapter (SureTrak II Universal Tracker; Medtronic) was installed on the handle of the sagittal surgical saw (GB129R; Aesculap, Pennsylvania, PA, USA) ( Fig. 2 ). Image-guided support was provided by the intraoperative navigation module of the MFSS, integrated with the infrared tracking camera of the commercial intraoperative navigation system (StealthStation S7; Medtronic). The registration process, based on six characteristic points marked as registration fiducials, was performed with a tip-pointer (Passive Planar Blunt Probe; Medtronic), according to the rigid-body point-based alignment of coordinate systems. After each registration procedure, the target fiducials in the area of interest, adjacent to the planned osteotomies, were used to determine TRE parameters, computed as the square root of the sum of squared deviation in all three spatial directions. The acquired mean registration procedure accuracy, expressed as a fiducial registration error (FRE), and the mean intraoperative navigation accuracy, expressed as TRE, were archived. An average FRE of <1.00 mm and an average TRE of <1.50 mm were considered indicative of a successfully conducted registration process.
Next, the calibration of the sagittal surgical saw blade was carried out with the use of a navigated pointer. During this procedure, the tooth edge width and length, along with the long axis plane of the blade were determined, allowing this tool to be navigated precisely. The position of the saw blade contours was shown in real-time on a screen in multiplanar 2D (axial, coronal, and frontal planes) and 3D views of the operative field. The accuracy of the saw blade navigation was evaluated by applying it to the target fiducial points of the mandible and comparing its position and angle in virtual and physical space. If the surgical instrument navigation was considered accurate, the calibration procedure was completed and the resection of bone structures was performed in accordance with the virtual surgery plan. Due to the image-guided support, the position and the tilt of the saw blade were displayed on the screen in real time in various 2D cross-sections (frontal, axial, and sagittal planes) and on the 3D image of the surgical field, which allowed the operator to perform the resection according to the planned osteotomy trajectory ( Fig. 2 ).
Postoperative CT scans, with identical imaging parameters, stored in DICOM file format, were obtained for all resected mandible segments. Using the MFSS software, postoperative image data were fused with the virtual preoperative CT-based plan, by labelling eight corresponding fiducial points. The accuracy of the fusion was calculated using the same formula as the TRE parameter and was measured in each case. An image fusion with average error of less than 1.00 mm was considered acceptable. With the image data superimposed, an objective, reliable measurement and analysis of the results of the osteotomy performed was possible. For each operation, the following parameters were analyzed twice by two observers: difference between the planned and actual volume of the resected bone fragment, angular deviation from the planned osteotomy trajectory ( Fig. 3 ), and differences in the positions of the marginal points labelled on the edges of the trajectory of the planned and actual osteotomy ( Fig. 4 ). The deviation in the location of these marginal points was calculated using the same formula as for the TRE parameter. A summary of all of the steps performed is provided in Fig. 5 .
The statistical analysis was performed using the two-tailed Student t -test. This was applied to verify the hypothesis that the two assessors would provide comparable measurements in terms of the mean value. It was also applied to compare the differences between the planned osteotomy and that performed. The intra-observer variability between the first and the second sets of measurements taken by the same assessor was assessed using the Bland–Altman method. The inter-observer variability between the two observers for the parameters measured in the first assessment series was also evaluated using the Bland–Altman method.
A total of 16 mandibular osteotomies (left ramus and angle, n = 4; right ramus and angle, n = 4; left body and mentum, n = 4; right body, n = 4) guided with MFSS were performed on four prefabricated models. The mean FRE was 0.70 ± 0.16 mm and the mean TRE was 0.95 ± 0.19 mm. The mean FREs and TREs of each type of procedure performed are listed in Table 1 . In all cases, sagittal surgical saw blade calibration was successful and took less than a minute.
Mean ± SD
Mean ± SD
|Left ramus + angle||0.66 ± 0.17||0.87 ± 0.23|
|Right ramus + angle||0.66 ± 0.10||1.02 ± 0.08|
|Left body + mentum||0.87 ± 0.03||1.01 ± 0.15|
|Right body||0.61 ± 0.14||0.90 ± 0.22|
As determined in the methodology, the objective analysis of results was performed twice by two observers. All postoperative image data were successfully fused with the virtual plan, with a mean fusion accuracy of 0.50 ± 0.10 mm. As stated, the two-tailed Student t -test was used to determine variation in inter-observer measurements. There was no statistically significant difference between the measurements obtained by the first and second evaluators. Therefore, all measurements of each type were combined and presented as the mean ± standard deviation (SD), unless stated otherwise. The analysis showed highly significant intra-observer repeatability for both assessors and a high level of inter-observer repeatability between the two assessors for the repeated measurements ( Table 2 ).
|Parameter type evaluated||Intra-observer variability||Inter-observer variability||P -value a|
|Observer A||Observer B|
|Fusion accuracy (mm)||0.12||0.14||0.12||0.93|
|Volume difference (cm 3 )||0.28||0.28||0.18||0.86|
|Angular deviation (°)||1.44||1.65||1.47||0.88|
|Marginal point deviation (mm)||0.34||0.33||0.32||0.91|
The measured differences between the virtual plans and the postoperative results are shown in Tables 3–5 . For all model osteotomies performed, the mean difference between the planned and actual bone resection volumes was 8.55 ± 5.51%, the mean angular deviation between the planned and executed osteotomy trajectory was 8.08 ± 5.50°, and the mean difference between the preoperative and postoperative marginal point positions amounted to 2.63 ± 1.27 mm. The Student t -test ( α = 1%) for discrepancies between the planned osteotomies and those performed indicated statistically significant differences for all angular parameters, while for the volume parameters, this applied to all procedures with the exception of ‘right body osteotomy’, for which the null hypothesis could not be rejected ( P = 0.13).
|Operation||Target volume (cm 3 )||Postoperative difference (cm 3 )
Mean ± SD
|Postoperative difference (%)
Mean ± SD
|Left ramus + angle||8.39||−0.18 ± 0.46||5.07 ± 2.81|
|Right ramus + angle||5.62||0.32 ± 0.23||5.73 ± 3.89|
|Left body + mentum||14.07||2.37 ± 0.25||16.81 ± 1.70|
|Right body||8.82||0.58 ± 0.16||6.61 ± 1.88|
|Operation||Proximal osteotomy angular deviation (°)
Mean ± SD
|Distal osteotomy angular deviation (°)
Mean ± SD
|Sagittal plane||Axial plane||Frontal plane||Sagittal plane||Axial plane||Frontal plane|
|Left ramus + angle||5.48 ± 4.40||5.56 ± 2.00||10.91 ± 6.67||9.05 ± 6.36|
|Right ramus + angle||7.05 ± 2.12||9.49 ± 6.00||2.01 ± 1.10||7.34 ± 1.64|
|Left body + mentum||3.37 ± 2.68||11.81 ± 3.53||7.47 ± 3.83||4.07 ± 1.97|
|Right body||3.34 ± 1.39||18.64 ± 1.38||11.28 ± 1.76||12.4 ± 2.75|