In order to assess 3D movement of bone segments between two different sets of volumetric data under separate conditions, accurate registration of the identical bone segments is required. In this study, we measured the DICE coefficient s using the overlapping volume of the segmented models as an index of segmentation quality. Since s was larger than 0.96, it is considered that the bone segments used in this study were segmented accurately from each set of data from the 6 patients. We believe that CA was extracted sufficiently because the CA contrast was clearly visible on the CTA image. We judged the anatomical validity of CA bifurcation points by referring to the contour of vasculature. However, these assessments could have measurement errors. In future studies, the accuracy and variability of segmentation of CA and identification of the bifurcations should be assessed so as to support the data obtained in this study more reliably.
The position and orientation parameters in CT volumetric data were represented in the initial DICOM coordinates (pre-treatment CTA coordinate system). Thus, the data are influenced by the relative position of the patient in the CT coordinate system. To accumulate displacements data under different imaging conditions, the obtained data should be transferred to the standardized coordinate system. Therefore, we set the coordinate system determined by the principal axes of inertia in the individual maxillofacial models.
CA deformation is obviously caused by changes in the head and neck position. In the present study, CA deformation was actually assessed using CTA data scanned in the supine position before and after treatment of oral cancer. Some CA bifurcation points shifted by more than 1 cm. This implies that preoperative data-based catheter navigation can include substantial positioning errors because of CA deformation. Additionally, in discussions on the accuracy of EM tracking systems [15, 16], CA deformation was considered to be more important than the registration or calibration procedure as a factor affecting errors in catheter navigation. In this study, the range of CA deformation was assessed at 5 important CA bifurcations for catheterization. When patients 2, 3, and 6 with small centroid shifts in the cervical vertebrae were defined as group A, and patients 1, 4, and 5 with large centroid shifts of the cervical vertebrae were defined as group B, the average centroid shifts of the C1–C4 models in groups A and B were 1.17 mm and 5.51 mm, respectively. As a result, the average shifts of CA bifurcation in group A and B were 2.51 mm and 5.84 mm, respectively. CA deformation was considered to be small when centroid movement in the surrounding bone models was small. The influence of mandibular displacement on CA deformation was not clear, because there were no significant mandibular displacements in 5 patients because of a stable occlusal position.
The correlation between the displacements of the surrounding bones and CA deformation was analyzed. C4 is most distal from the maxillofacial bone. C2 is mainly moved by motion of the atlantoaxial joint during rotation of the neck. Thus, C4 and C2 were selected for analysis. To represent rotation and anteflexion of the head, differences between translations and rotations of C4 and C2 were used. The following parameters were calculated: x
1 (C4Δx´ − C2Δx´), x
2 (C4Δy´ − C2Δy´), x
3 (C4Δz´ − C2Δz´), and x
4 (C4Δα´ − C2Δα´) and selected as 4 parameters representing the displacements of surrounding bones caused by head rotation and anteflexion. These 4 parameters were fitted to the displacement of CA bifurcation points by linear multiple regression analysis.
That is,
(1)
b
0
was a constant and b
1
, b
2
, b
3
, b
4
were partial regression coefficients. The relationship between the calculated values using the multiple regression model and the observed values is indicated in Figure 7. We suggest the possibility that the displacement (position and orientation) of surrounding bones is related to the displacement (position) of CA bifurcations. Although angiographic data of patients’ non-catheterized CAs (left or right) are rare because of the clinical protocol used for CA catheterization, analysis data for a larger number of cases will be required to confirm this hypothesis.
For the clinical application of catheter navigation, a method to minimize the shift in the pre- and intra-operative positions of the mandible and of the cervical vertebrae is required. To prevent this shift, a facial mask can be used as a non-invasive immobilization system [17, 18]. Furthermore, invasive methods are available to fix the head using head pins. The shift in the CA bifurcation was approximately 2–3 mm without large movement of the cervical vertebrae and mandible although this result should have measurement errors.
To improve accuracy of catheter navigation, it is important to develop a method to predict shifts in CA bifurcations because of intraoperative deformation. The present study indicated the importance of the movement of surrounding bones in predicting CA bifurcation. The position and orientation of the surrounding bones can be measured using intraoperative X-ray images, and the intraoperative position and orientation can be compared with the preoperative ones. By investigating the correlation between the displacements of surrounding bones and CA bifurcation using accumulated clinical data, we will be able to develop a method to predict CA bifurcation points without a contrast medium for fluoroscopic images.
In addition, a registration method by which the path of a sensor attached to an intravascular catheter corresponds to the artery centerline in CTA images is required. The registration method between the intravascular path and the vessel centerline in the preoperative images has been reported in cerebrovascular surgery [19], bronchoscopy [20], and cardiovascular surgery [21]. Our research group has also reported its possible application in the head and neck [5]. In future studies, the results of the present study will provide the reference data.