Abstract
The accuracy of three-dimensional (3D) predictions of soft tissue changes in the surgical correction of facial asymmetry was evaluated in this study. Preoperative (T1) and 6–12-month postoperative (T2) cone beam computed tomography scans of 13 patients were studied. All patients underwent surgical correction of facial asymmetry as part of a multidisciplinary treatment protocol. The magnitude of the surgical movement was measured; virtual surgery was performed on the preoperative scans using Maxilim software. The predicted soft tissue changes were compared to the actual postoperative appearance (T2). Mean (signed) distances and mean (absolute) distances between the predicted and actual 3D surface meshes for each region were calculated. The one-sample t -test was applied to test the alternative hypothesis that the mean absolute distances had a value of <2.0 mm. A novel directional analysis was applied to analyse the accuracy of the prediction of soft tissue changes. The results showed that the distances between the predicted and actual postoperative soft tissue changes were less than 2.0 mm in all regions. The predicted facial morphology was narrower than the actual surgical changes in the cheek regions. 3D soft tissue prediction using Maxilim software in patients undergoing the correction of facial asymmetry is clinically acceptable.
Facial asymmetry poses a challenge in craniofacial diagnosis and treatment planning. An appropriate assessment and quantification of the differences between the right and left sides of the face is essential for diagnosis, treatment planning, and follow-up. Recent advancements made in three-dimensional (3D) imaging and image analysis offer the potential for developments in the field of facial deformity diagnosis. The advent of cone beam computed tomography (CBCT) has allowed the examination of anatomical structures in multiplanar views, which has led to the more accurate and comprehensive diagnosis of facial asymmetry.
To obtain the ideal facial aesthetics and functional results in orthognathic surgery, accurate planning is crucial. This requires accurate prediction and a comprehensive understanding of the soft tissue response to the underlying surgical movement of the jaw bones . Knowledge of the soft tissue response not only helps to guide the surgical movements of the osteotomy segments, but also informs the need for preparative orthodontic ‘decompensation’ to achieve the required surgical skeletal correction.
Prediction planning in the surgical correction of facial asymmetry is essential to maximize the harmony of facial aesthetics following orthognathic surgery . It also plays an important role in communication between the surgeon, the orthodontist, and the patient . With realistic expectations of the results of surgery for the correction of facial asymmetry, patient dissatisfaction with the surgical outcome will be minimized .
Maxilim software (Medicim – Medical Image Computing, Mechelen, Belgium) uses the mass tensor model algorithm. This software has been validated in a study of CT data of 10 patients with combined dentoskeletal deformities . In that study, the greatest areas of surgical change were observed at the lip and chin regions. The average mean distance between the real changes and the predicted changes of the soft tissue in response to orthognathic surgery was 0.6 mm and the average 90 th percentile was below 1.5 mm .
Shafi et al. validated the accuracy of Maxilim software in predicting soft tissue changes following Le Fort I maxillary advancement surgery . Their study included 13 patients, and the average surgical advancement was 5.5 ± 2.2 mm. The accuracy of prediction was less than 3 mm in the majority of the facial regions. However, the prediction error in the upper lip region was more than 3 mm.
Nadjmi et al. showed that the mean absolute distance between the predicted and the actual postoperative soft tissue changes was 1.18 mm, with the lowest reported accuracy in the lower lip region . Prediction errors in the cheek region were observed, especially in cases of vertical maxillary excess where the upper jaw impaction exceeded 4 mm.
The accuracy of Maxilim software in predicting the soft tissue changes following bilateral sagittal split advancement osteotomies has also been investigated. In a study involving 100 cases by Liebregts et al. , the mean absolute prediction error for the entire face was 0.9 ± 0.3 mm and the mean absolute 90 th percentile was 1.9 mm. The absolute mean prediction error was less than 2 mm in all of the facial tissue regions except the lower lip region. The lower lip area showed the lowest prediction accuracy, with a mean absolute error of 1.2 ± 0.5 mm. In bimaxillary surgeries, the accuracy of Maxilim software was reported to be 0.8 mm for the full face, 1.2 mm for the upper lip, 1.4 mm for the lower lip, and 1.1 mm for the chin area. The mean absolute error at the 90 th percentile of the corresponding 3D facial surface meshes was found to be less than 2 mm. The mean absolute prediction error for the whole face was less than 0.8 mm . Of note, the surgical correction in this study group was less than that in the research conducted by Shafi et al. .
Various algorithms have been developed and incorporated into commercially available prediction software packages to quantify facial soft tissue changes in three dimensions. These include the mass spring model, finite element model, and mass tensor model. Therefore, it would be prudent to assess the accuracy of soft tissue prediction, especially in the surgical correction of facial asymmetry cases where two-dimensional (2D) prediction planning is of limited value. This has not been investigated before.
The aim of this study was to evaluate the accuracy of 3D predictions of soft tissue changes following orthognathic surgery for the correction of facial asymmetry.
Materials and methods
The study sample comprised 13 non-syndromic adults with facial asymmetry, with a midline deviation of the chin point not less than 2.0 mm. All patients underwent single-jaw or bimaxillary osteotomies for the correction of facial asymmetry. The patients were examined at the same multidisciplinary clinic. CBCT scans were captured before surgical correction (T1) and at 6 to 12 months postoperative (T2) using an iCAT scanner with an isotropic voxel size resolution of 0.4 mm (Imaging Sciences, Hatfield, PA, USA).
Quantifying surgical skeletal movements
The images obtained at T1 and T2 were superimposed on the anterior cranial base, which was unaffected by the surgical procedure. OnDemand3D software (version 1.0; Cybermed, Seoul, South Korea) was used for voxel-based registration of the DICOM images of T2 to T1 in each case. Following a preliminary manual alignment of the images, voxel-based registration using OnDemand3D was performed to match the selected regions on the two DICOM Images . The registration was performed at the base of the skull which was not affected by surgery. This method maximizes the accuracy of the superimposition of the two images based on the grey-scale intensity, voxel by voxel. The superimposed DICOM images were transferred to Maxilim software (Medicim – Medical Image Computing).
The anteroposterior, vertical, and mediolateral surgical movements of the osteotomy segments of the maxilla and mandible were measured using a novel method, which was developed by our research team .
Prediction of soft tissue changes
The prediction of soft tissue changes was conducted on all T1 images using Maxilim software guided by the measured actual surgical movements, which were identified as explained above. The soft tissue simulation was performed according to the manufacturer’s instructions. The prediction of soft tissue changes using Maxilim software is based on the mass tensor algorithm. Soft tissue surface models of both the simulated and the actual postoperative images were generated and exported in stereolithography (STL) file format for further analysis.
Assessment of the accuracy of the prediction planning
The surface differences between the predicted and the actual soft tissue changes were measured using VRMesh software (VirtualGrid, Bellevue City, WA, USA). A colour-coded distance map was generated to illustrate the magnitude and anatomical regions with prediction inaccuracies. The 3D surface model was segmented into the following anatomical regions: upper lip, lower lip, chin area, right and left paranasal regions, nose, and right and left cheeks. The reproducibility of the segmentation depends on the accuracy of the repeated digitization of facial landmarks, which was investigated in this study. The methodology has been validated by our research team . All of the models were converted into VRML (Virtual Reality Modelling Language) file format and were exported for further analysis. The surface distances between the predicted facial regions and those of the actual postoperative model were measured using in-house software developed for this purpose. Using this software, the minimum, maximum, mean, standard deviation (SD), absolute maximum, absolute mean, and absolute SD of 90% of the points of each facial anatomical region were measured.
Further, in order to understand the directional accuracy of the prediction of the soft tissue changes in response to surgery, a generic model was conformed to the predicted models and the postoperative models to establish surface correspondences, as suggested by Cheung et al. . A generic 3D facial mesh comprises 2000 vertices. The vertices of the generic mesh are indexed so that the magnitude and the direction of discrepancies between corresponding 3D surfaces can be measured accurately. The generic mesh underwent conformation (elastic deformation) to resemble both the predicted and the actual postoperative 3D facial morphology of each patient. The prediction discrepancies at each vertex were analysed in the x , y , and z dimensions separately. The results of the differences in each direction were displayed in colour-coded facial maps. The conformation of a generic surface mesh on a 3D facial surface has an acceptable level of accuracy and has been validated for measuring facial changes between images, i.e. pre- and post-surgery . Areas of over-prediction and under-prediction of soft tissues changes in response to the correction of facial asymmetry were identified by measuring the mean corresponding differences between the predicted models and the postoperative models.
The one-sample t -test (one-tailed) was applied to the mean absolute distances, with the alternative hypothesis that assumed the true mean of the sample to be less than 2 mm. A significance level of 0.05 was used for the test. To evaluate the validity and reproducibility of the study, the entire measurement procedure was performed twice with a 1-week interval.
Results
The one-sample Kolmogorov–Smirnov test confirmed that the data were normally distributed before the t -test was applied.
Distances between the predicted and the actual soft tissue changes
The mean surgical correction of the asymmetry in the pogonion region was 3.6 mm (SD 4.2 mm). Ten of the 13 patients underwent bimaxillary surgery (Le Fort I osteotomy and bilateral sagittal split osteotomy); a condylectomy was also performed in one of these patients. A bilateral sagittal split mandibular osteotomy was performed in the remaining three cases, with an additional genioplasty performed in one of them. Alar cinch suture and V–Y closure were applied for all patients who underwent Le Fort I osteotomy.
No statistically significant differences between the repeated facial measurements were found ( Table 1 ), and the intra-class correlation (ICC) values confirmed that the facial measurements are reproducible.
| Regions | Anatomical region | Error of signed distances (mm) | SD of the error of signed distances | Error of absolute distances (mm) | SD of the error of absolute distances | P -value for the repeated mean absolute distances | P -value for the repeated mean signed distances | Intra-class correlation |
|---|---|---|---|---|---|---|---|---|
| 1 | Upper lip | 0.05 | 0.47 | 0.03 | 0.46 | 0.82 | 0.74 | 0.83 |
| 2 | Lower lip | −0.35 | 1.29 | −0.03 | 0.20 | 0.65 | 0.35 | 0.64 |
| 3 | Chin | −0.24 | 0.66 | 0.1 | 0.22 | 0.13 | 0.22 | 0.70 |
| 4 | R paranasal region | 0.24 | 0.98 | 0.00 | 0.32 | 0.97 | 0.41 | 0.65 |
| 5 | L paranasal region | 0.10 | 0.86 | −0.07 | 0.24 | 0.31 | 0.68 | 0.86 |
| 6 | Nose | −0.24 | 0.45 | 0.00 | 0.04 | 0.89 | 0.08 | 0.60 |
| 7 | L cheek | 0.49 | 1.26 | 0.14 | 0.39 | 0.23 | 0.18 | 0.62 |
| 8 | R cheek | −0.22 | 0.72 | 0.08 | 0.59 | 0.62 | 0.30 | 0.88 |
Table 2 shows the mean surface differences (mm) between the predicted and actual surface meshes of the facial soft tissue regions. The results showed a tendency towards under-prediction of the soft tissue changes in response to the correction of facial asymmetry in all of the anatomical regions except the upper lip and the left cheek, where a tendency of over-prediction was recorded.
| Regions | Anatomical region | Mean of the mean distances (mm) | Mean of the mean absolute distances (mm) | SD of the absolute distances | P -value (one-sample t -test) | Percentage of the mean absolute distances <2 mm |
|---|---|---|---|---|---|---|
| 1 | Upper lip | 0.43 | 0.86 | 0.43 | 2.69 × 10 −7 | 100% |
| 2 | Lower lip | 0.20 | 0.97 | 0.40 | 4.20 × 10 −7 | 100% |
| 3 | Chin | −0.39 | 0.77 | 0.43 | 1.45 × 10 −7 | 100% |
| 4 | R paranasal region | −0.37 | 0.96 | 0.52 | 5.30 × 10 −6 | 92.3% |
| 5 | L paranasal region | −0.55 | 0.96 | 0.61 | 2.73 × 10 −5 | 92.3% |
| 6 | Nose | −0.40 | 0.60 | 0.25 | 6.76 × 10 −11 | 100% |
| 7 | L cheek | −0.30 | 1.30 | 0.62 | 8.35 × 10 −4 | 76.9% |
| 8 | R cheek | 0.24 | 1.08 | 0.55 | 3.00 × 10 −5 | 92.3% |
Stay updated, free dental videos. Join our Telegram channel
VIDEdental - Online dental courses
