Improving Image Segmentation with Contextual and Structural Similarity.

Deep learning models for medical image segmentation are usually trained with voxel-wise losses, e.g., cross-entropy loss, focusing on unary supervision without considering inter-voxel relationships. This oversight potentially leads to semantically inconsistent predictions. Here, we propose a contextual similarity loss (CSL) and a structural similarity loss (SSL) to explicitly and efficiently incorporate inter-voxel relationships for improved performance. The CSL promotes consistency in predicted object categories for each image sub-region compared to ground truth. The SSL enforces compatibility between the predictions of voxel pairs by computing pair-wise distances between them, ensuring that voxels of the same class are close together whereas those from different classes are separated by a wide margin in the distribution space. The effectiveness of the CSL and SSL is evaluated using a clinical cone-beam computed tomography (CBCT) dataset of patients with various craniomaxillofacial (CMF) deformities and a public pancreas dataset. Experimental results show that the CSL and SSL outperform state-of-the-art regional loss functions in preserving segmentation semantics.

Midsagittal Plane First: Building a Strong Facial Reference Frame for Computer-Aided Surgical Simulation.

PURPOSE: A facial reference frame is a 3-dimensional Cartesian coordinate system that includes 3 perpendicular planes: midsagittal, axial, and coronal. The order in which one defines the planes matters. The purposes of this study are to determine the following: 1) what sequence (axial-midsagittal-coronal vs midsagittal-axial-coronal) produced more appropriate reference frames and 2) whether orbital or auricular dystopia influenced the outcomes. METHODS: This study is an ambispective cross-sectional study. Fifty-four subjects with facial asymmetry were included. The facial reference frames of each subject (outcome variable) were constructed using 2 methods (independent variable): axial plane first and midsagittal plane first. Two board-certified orthodontists together blindly evaluated the results using a 3-point categorical scale based on their careful inspection and expert intuition. The covariant for stratification was the existence of orbital or auricular dystopia. Finally, Wilcoxon signed rank tests were performed. RESULTS: The facial reference frames defined by the midsagittal plane first method was statistically significantly different from ones defined by the axial plane first method (P = .001). Using the midsagittal plane first method, the reference frames were more appropriately defined in 22 (40.7%) subjects, equivalent in 26 (48.1%) and less appropriately defined in 6 (11.1%). After stratified by orbital or auricular dystopia, the results also showed that the reference frame computed using midsagittal plane first method was statistically significantly more appropriate in both subject groups regardless of the existence of orbital or auricular dystopia (27 with orbital or auricular dystopia and 27 without, both P < .05). CONCLUSIONS: The midsagittal plane first sequence improves the facial reference frames compared with the traditional axial plane first approach. However, regardless of the sequence used, clinicians need to judge the correctness of the reference frame before diagnosis or surgical planning.

A Better Understanding of Unilateral Condylar Hyperplasia of the Mandible.

PURPOSE: Our current understanding of unilateral condylar hyperplasia (UCH) was put forth by Obwegeser. He hypothesized that UCH is 2 separate conditions: hemimandibular hyperplasia and hemimandibular elongation. This hypothesis was based on the following 3 assumptions: 1) the direction of overgrowth, in UCH, is bimodal-vertical or horizontal, with rare cases growing obliquely; 2) UCH can expand a hemimandible with and without significant condylar enlargement; and 3) there is an association between the condylar expansion and the direction of overgrowth-minimal expansion resulting in horizontal growth and significant enlargement causing vertical displacement. The purpose of this study was to test these assumptions. PATIENTS AND METHODS: We analyzed the computed tomography scans of 40 patients with UCH. First, we used a Silverman Cluster analysis to determine how the direction of overgrowth is distributed in the UCH population. Next, we evaluated the relationship between hemimandibular overgrowth and condylar enlargement to confirm that overgrowth can occur independently of condylar expansion. Finally, we assessed the relationship between the degree of condylar enlargement and the direction of overgrowth to ascertain if condylar expansion determines the direction of growth. RESULTS: Our first investigation demonstrates that the general impression that UCH is bimodal is wrong. The growth vectors in UCH are unimodally distributed, with the vast majority of cases growing diagonally. Our second investigation confirms the observation that UCH can expand a hemimandible with and without significant condylar enlargement. Our last investigation determined that in UCH, there is no association between the degree of condylar expansion and the direction of the overgrowth. CONCLUSIONS: The results of this study disprove the idea that UCH is 2 different conditions: hemimandibular hyperplasia and hemimandibular elongation. It also provides new insights about the pathophysiology of UCH.

Validity of Medical Insurance Guidelines for Orthognathic Surgery.

PURPOSE: The purpose of this study was to assess the validity of the medical insurance guidelines for orthognathic surgery used by the major American medical insurance companies. MATERIALS AND METHODS: This study assessed the validity of the medical insurance guidelines for orthognathic surgery used by Aetna, Anthem Blue Cross Blue Shield (BCBS), Cigna, Humana, and UnitedHealthcare (UHC). To evaluate the validity, we calculated the approval and denial rates of the 5 guidelines when we used them to assess the medical necessity for a control group of carefully selected patients. Patients were included in the control group if they met the criteria of a "prudent provider," crafted for this study. All rejected cases were analyzed to determine the root cause of the denials. The validity of the guidelines was also ascertained by determining their completeness and correctness. RESULTS: The current study proves that no insurance guideline is in agreement with the criteria of a "prudent provider." When applied to carefully chosen patients, the requirements of BCBS, Aetna, Humana, and Cigna produce modest rejection rates of 6 to 12%. UHC is an outlier. Its guideline rejects 86% of patients, a rate about 7 times higher than its peers. Insurance guidelines disqualified patients for 3 different reasons: 1) no significant jaw deformity, 2) no demonstrable health impairment, and 3) the etiology of the condition is not a covered benefit. Additional evaluations demonstrate that the private insurance guidelines are incomplete, and at times, incorrect. CONCLUSIONS: This study shows that the medical insurance guidelines for orthognathic surgery used by the major American medical insurance plans need revision. The most consequential flaw was considering etiology in judging medical necessity. Fortunately, only one company adopted this policy. Moreover, all guidelines have omissions and errors in the way jaw deformity is determined and how health impairment is determined.

Clinical Evaluation of Digital Dental Articulation for One-Piece Maxillary Surgery.

PURPOSE: Methods for digital dental alignment are not readily available to automatically articulate the upper and lower jaw models. The purpose of the present study was to assess the accuracy of our newly developed 3-stage automatic digital articulation approach by comparing it with the reference standard of orthodontist-articulated occlusion. MATERIALS AND METHODS: Thirty pairs of stone dental models from double-jaw orthognathic surgery patients who had undergone 1-piece Le Fort I osteotomy were used. Two experienced orthodontists manually articulated the models to their perceived final occlusion for surgery. Each pair of models was then scanned twice-while in the orthodontist-determined occlusion and again with the upper and lower models separated and positioned randomly. The separately scanned models were automatically articulated to the final occlusion using our 3-stage algorithm, resulting in an algorithm-articulated occlusion (experimental group). The models scanned together represented the manually articulated occlusion (control group). A qualitative evaluation was completed using a 3-point categorical scale by the same orthodontists, who were unaware of the methods used to articulate the models. A quantitative evaluation was also completed to determine whether any differences were present in the midline, canine, and molar relationships between the algorithm-determined and manually articulated occlusions using repeated measures analysis of variance (ANOVA). Finally, the mean ± standard deviation values were computed to determine the differences between the 2 methods. RESULTS: The results of the qualitative evaluation revealed that all the algorithm-articulated occlusions were as good as the manually articulated ones. The results of the repeated measures ANOVA found no statistically significant differences between the 2 methods [F(1,28) = 0.03; P = .87]. The mean differences between the 2 methods were all within 0.2 mm. CONCLUSIONS: The results of our study have demonstrated that dental models can be accurately, reliably, and automatically articulated using our 3-stage algorithm approach, meeting the reference standard of orthodontist-articulated occlusion.

Both the Observer's Expertise and the Subject's Facial Symmetry Can Affect Anatomical Position of the Head.

PURPOSE: It is easier to judge facial deformity when the patient's head is in anatomic position. The purposes of this study were to determine 1) whether a group of expert observers would agree more than a group of nonexperts on what is the correct anatomic position of the head, 2) whether there would be more variation in the alignment of an asymmetrical face compared with a symmetrical one, and 3) whether the alignments of experts would be more repeatable than those of nonexperts. MATERIALS AND METHODS: Thirty-one orthodontists (experts) and 31 dental students (nonexperts) were recruited for this mixed-model study. They were shown randomly oriented 3-dimensional head photographs of an adult with a symmetrical face and an adolescent with an asymmetrical face. In viewing software, the observers oriented the images into anatomic position. They repeated the orientations 4 weeks later. Data were analyzed using a generalized linear model and Bland-Altman plots. The primary predictor variables were experience and symmetry status. The outcome variable was the anatomic position of the head. The other variables of interest included time and orientation direction. RESULTS: There was a statistically significant difference between measurements completed by experts and nonexperts (F1,60 = 14.83; P < .01). The interaction between expertise and symmetrical status showed a statistically significant difference between symmetrical and asymmetrical faces in the expert and nonexpert groups (F1,60 = 9.93; P = .003). The interaction between expertise and time showed a statistically significant difference in measurement over time in the expert and nonexpert groups (F1,60 = 4.66; P = .03). CONCLUSIONS: The study shows that experts can set a head into anatomic position better than nonexperts. In addition, facial asymmetry has a profound effect on the ability of an observer to align a head in the correct anatomic position. Moreover, observer-guided alignment is not reproducible.

A Modified Method of Proximal Segment Alignment After Sagittal Split Ramus Osteotomy for Patients With Mandibular Asymmetry.

PURPOSE: The purpose of this study was to evaluate a modified method of aligning the proximal segment after bilateral sagittal split ramus osteotomy (BSSO) in the treatment of patients with facial asymmetry. PATIENTS AND METHODS: Eleven patients with mandibular excess and facial asymmetries were enrolled in this prospective study. The surgery was planned according to a computer-aided surgical simulation protocol. In addition, the proximal segment on the hypoplastic side was intentionally flared out after the distal segment was rotationally set back. If the gap between the proximal and distal segments was too wide, then bone grafts were used. The surgery was completed according to the computerized plan. The proximal segment on the hypoplastic side was fixed with bicortical lag screws, and the proximal segment on the hyperplastic side was fixed with a 4-hole titanium miniplate. Postoperative evaluation was performed 6 months after surgery. Statistical analyses were performed. RESULTS: All surgeries were completed uneventfully. Of the 11 patients, 4 also underwent genioplasty and 3 underwent bone grafting to fill in the gap and smooth the anterior step. The physicians and patients were satisfied with the surgical outcomes. Only 1 patient underwent a secondary revision using an onlay hydroxyapatite implant. Results of statistical analyses showed that the computerized surgical plan could be accurately transferred to the patients at the time of surgery and the surgical outcomes achieved with this modified method were better than with the routine method of aligning the proximal and distal segments in maximal contact. CONCLUSION: The present modified method of aligning the proximal segment for BSSO can effectively correct mandibular asymmetry and obviate a secondary revision surgery.

Microscopic versus open approach to craniosynostosis: a long-term outcomes comparison.

The purpose of this retrospective study was to evaluate the long-term outcomes of using the microscopic minimally invasive approach for the treatment of nonsyndromic craniosynostosis. During the last 10 years, 180 consecutive patients with nonsyndromic craniosynostosis were treated: 67 patients were treated with microscopic minimally invasive approach, and 113 were treated with the open approach. In the microscopic group, there was 1 intraoperative complication (1.5%). There were 10 postoperative complications (14.9%), of which 9 required major reoperations and 1 required a minor procedure. The major complications occurred in 7 unicoronal patients (58.3%) and 2 metopic patients (25.0%). In the open-approach group, there were 8 complications (7.1%), 2 patients required major reoperations and 6 required minor procedures. Chi-squared test showed that there was no statistically significant difference in the overall complication rate between the microscopic and open approaches. However, in the unicoronal patients, the complication rate was significantly higher in the microscopic group (P < 0.001). In conclusion, the microscopic approach is our treatment of choice in nonsyndromic patients with sagittal and lambdoidal craniosynostosis. We no longer use the microscopic approach in patients with unicoronal or metopic craniosynostosis because of the high complication rate.

Accuracy of Surgical Outcome Using Computer-Aided Surgical Simulation in Fronto-Orbital Advancement for Craniosynostosis: A Pilot Study.

BACKGROUND AND OBJECTIVE: Computer-aided surgical simulation (CASS) can be used to virtually plan ideal outcomes of craniosynostosis surgery. Our purpose was to create a workflow analyzing the accuracy of surgical outcomes relative to virtually planned fronto-orbital advancement (FOA). METHODS: Patients who underwent FOA using CASS between October 1, 2017, and February 28, 2022, at our center and had postoperative computed tomography within 6 months of surgery were included. Virtual 3-dimensional (3D) models were created and coregistered using each patient's preoperative and postoperative computed tomography data. Three points on each bony segment were used to define the object in 3D space. Each planned bony segment was manipulated to match the actual postoperative outcome. The change in position of the 3D object was measured in translational (X, Y, Z) and rotational (roll, pitch, yaw) aspects to represent differences between planned and actual postoperative positions. The difference in the translational position of several bony landmarks was also recorded. Wilcoxon signed-rank tests were performed to measure significance of these differences from the ideal value of 0, which would indicate no difference between preoperative plan and postoperative outcome. RESULTS: Data for 63 bony segments were analyzed from 8 patients who met the inclusion criteria. Median differences between planned and actual outcomes of the segment groups ranged from -0.3 to -1.3 mm in the X plane; 1.4 to 5.6 mm in the Y plane; 0.9 to 2.7 mm in the Z plane; -1.2° to -4.5° in pitch; -0.1° to 1.0° in roll; and -2.8° to 1.0° in yaw. No significant difference from 0 was found in 21 of 24 segment region/side combinations. Translational differences of bony landmarks ranged from -2.7 to 3.6 mm. CONCLUSION: A high degree of accuracy was observed relative to the CASS plan. Virtual analysis of surgical accuracy in FOA using CASS was feasible.

Correspondence attention for facial appearance simulation.

In orthognathic surgical planning for patients with jaw deformities, it is crucial to accurately simulate the changes in facial appearance that follow the bony movement. Compared with the traditional biomechanics-based methods like the finite-element method (FEM), which are both labor-intensive and computationally inefficient, deep learning-based methods offer an efficient and robust modeling alternative. However, current methods do not account for the physical relationship between facial soft tissue and bony structure, causing them to fall short in accuracy compared to FEM. In this work, we propose an Attentive Correspondence assisted Movement Transformation network (ACMT-Net) to predict facial changes by correlating facial soft tissue changes with bony movement through a point-to-point attentive correspondence matrix. To ensure efficient training, we also introduce a contrastive loss for self-supervised pre-training of the ACMT-Net with a k-Nearest Neighbors (k-NN) based clustering. Experimental results on patients with jaw deformities show that our proposed solution can achieve significantly improved computational efficiency over the state-of-the-art FEM-based method with comparable facial change prediction accuracy.

Accuracy of a computer-aided surgical simulation protocol for orthognathic surgery: a prospective multicenter study.

PURPOSE: The purpose of this prospective multicenter study was to assess the accuracy of a computer-aided surgical simulation (CASS) protocol for orthognathic surgery. MATERIALS AND METHODS: The accuracy of the CASS protocol was assessed by comparing planned outcomes with postoperative outcomes of 65 consecutive patients enrolled from 3 centers. Computer-generated surgical splints were used for all patients. For the genioplasty, 1 center used computer-generated chin templates to reposition the chin segment only for patients with asymmetry. Standard intraoperative measurements were used without the chin templates for the remaining patients. The primary outcome measurements were the linear and angular differences for the maxilla, mandible, and chin when the planned and postoperative models were registered at the cranium. The secondary outcome measurements were the maxillary dental midline difference between the planned and postoperative positions and the linear and angular differences of the chin segment between the groups with and without the use of the template. The latter were measured when the planned and postoperative models were registered at the mandibular body. Statistical analyses were performed, and the accuracy was reported using root mean square deviation (RMSD) and the Bland-Altman method for assessing measurement agreement. RESULTS: In the primary outcome measurements, there was no statistically significant difference among the 3 centers for the maxilla and mandible. The largest RMSDs were 1.0 mm and 1.5° for the maxilla and 1.1 mm and 1.8° for the mandible. For the chin, there was a statistically significant difference between the groups with and without the use of the chin template. The chin template group showed excellent accuracy, with the largest positional RMSD of 1.0 mm and the largest orientation RMSD of 2.2°. However, larger variances were observed in the group not using the chin template. This was significant in the anteroposterior and superoinferior directions and the in pitch and yaw orientations. In the secondary outcome measurements, the RMSD of the maxillary dental midline positions was 0.9 mm. When registered at the body of the mandible, the linear and angular differences of the chin segment between the groups with and without the use of the chin template were consistent with the results found in the primary outcome measurements. CONCLUSIONS: Using this computer-aided surgical simulation protocol, the computerized plan can be transferred accurately and consistently to the patient to position the maxilla and mandible at the time of surgery. The computer-generated chin template provides greater accuracy in repositioning the chin segment than the intraoperative measurements.

Biomechanical evaluation of a new MatrixMandible plating system on cadaver mandibles.

PURPOSE: Current mandibular plating systems contain a wide range of plates and screws needed for the treatment of mandibular reconstruction and mandibular fractures. The authors' hypothesis was that a single diameter screw could be used in all applications in a plating system. Therefore, the purpose of this study was to test if the 2.0-mm locking screws could replace the 2.4-mm screws to stabilize a 2.5-mm-thick reconstruction plate in the treatment of mandibular discontinuity. MATERIALS AND METHODS: Thirty-six fresh human cadaveric mandibles were used: 18 were plated using 2.0-mm locking screws (experimental) and the other 18 were plated using 2.4-mm locking screws (control). Each group was further divided into 3 subgroups based on the site of loading application: the ipsilateral (right) second premolar region, the central incisal region, and the contralateral (left) first molar region. The same ipsilateral (right) mandibular angular discontinuity was created by the same surgeon. The mandible was mounted on a material testing machine. The micromotions between the 2 segments, permanent and elastic displacements, were recorded after incremental ramping loads. The magnitude of screw back-out and the separation between plate and bone were recorded using a laser scanner (resolution, 0.12 mm) before and after the loading applications. The data were processed. Descriptive analyses and a general linear model for repeated measures analysis of variance were performed. RESULTS: There was no statistically significant difference in permanent displacement (mean, 1.16 and 0.82 mm, respectively) between the 2.0-mm and 2.4-mm screw groups. There also was no statistically significant difference in elastic displacement between the 2 groups (mean, 1.48 and 1.21 mm, respectively). Finally, there were no statistically significant differences in screw back-out or separation between plate and bone between the 2 groups. All means for screw back-out and separation between screw and bone for each group were judged within the error of the laser scanning system (<0.12 mm). CONCLUSION: One may anticipate that the mechanical functions of the 2.0-mm locking screws are not different from those of the 2.4-mm screws when a 2.5-mm-thick reconstruction plate is used to reconstruct mandibular angular discontinuity. However, further biomechanical studies (ie, fatigue of screws) are warranted before a randomized clinical trial can be conducted to definitively prove that the 2.4-mm screws can be replaced by 2.0-mm screws.

3D segmentation of maxilla in cone-beam computed tomography imaging using base invariant wavelet active shape model on customized two-manifold topology.

Recent advances in cone-beam computed tomography (CBCT) have rapidly enabled widepsread applications of dentomaxillofacial imaging and orthodontic practices in the past decades due to its low radiation dose, high spatial resolution, and accessibility. However, low contrast resolution in CBCT image has become its major limitation in building skull models. Intensive hand-segmentation is usually required to reconstruct the skull models. One of the regions affected by this limitation the most is the thin bone images. This paper presents a novel segmentation approach based on wavelet density model (WDM) for a particular interest in the outer surface of anterior wall of maxilla. Nineteen CBCT datasets are used to conduct two experiments. This mode-based segmentation approach is validated and compared with three different segmentation approaches. The results show that the performance of this model-based segmentation approach is better than those of the other approaches. It can achieve 0.25 ± 0.2 mm of surface error from ground truth of bone surface.

Simulation of Postoperative Facial Appearances via Geometric Deep Learning for Efficient Orthognathic Surgical Planning.

Orthognathic surgery corrects jaw deformities to improve aesthetics and functions. Due to the complexity of the craniomaxillofacial (CMF) anatomy, orthognathic surgery requires precise surgical planning, which involves predicting postoperative changes in facial appearance. To this end, most conventional methods involve simulation with biomechanical modeling methods, which are labor intensive and computationally expensive. Here we introduce a learning-based framework to speed up the simulation of postoperative facial appearances. Specifically, we introduce a facial shape change prediction network (FSC-Net) to learn the nonlinear mapping from bony shape changes to facial shape changes. FSC-Net is a point transform network weakly-supervised by paired preoperative and postoperative data without point-wise correspondence. In FSC-Net, a distance-guided shape loss places more emphasis on the jaw region. A local point constraint loss restricts point displacements to preserve the topology and smoothness of the surface mesh after point transformation. Evaluation results indicate that FSC-Net achieves 15× speedup with accuracy comparable to a state-of-the-art (SOTA) finite-element modeling (FEM) method.

Bidirectional prediction of facial and bony shapes for orthognathic surgical planning.

This paper proposes a deep learning framework to encode subject-specific transformations between facial and bony shapes for orthognathic surgical planning. Our framework involves a bidirectional point-to-point convolutional network (P2P-Conv) to predict the transformations between facial and bony shapes. P2P-Conv is an extension of the state-of-the-art P2P-Net and leverages dynamic point-wise convolution (i.e., PointConv) to capture local-to-global spatial information. Data augmentation is carried out in the training of P2P-Conv with multiple point subsets from the facial and bony shapes. During inference, network outputs generated for multiple point subsets are combined into a dense transformation. Finally, non-rigid registration using the coherent point drift (CPD) algorithm is applied to generate surface meshes based on the predicted point sets. Experimental results on real-subject data demonstrate that our method substantially improves the prediction of facial and bony shapes over state-of-the-art methods.

In vitro evaluation of new approach to digital dental model articulation.

PURPOSE: The purpose of the present study was to evaluate the accuracy of our newly developed approach to digital dental model articulation. MATERIALS AND METHODS: Twelve sets of stone dental models from patients with craniomaxillofacial deformities were used for validation. All the models had stable occlusion and no evidence of early contact. The stone models were hand articulated to the maximal intercuspation (MI) position and scanned using a 3-dimensional surface laser scanner. These digital dental models at the MI position served as the control group. To establish an experimental group, each mandibular dental model was disarticulated from its original MI position to 80 initial positions. Using a regular office personal computer, they were digitally articulated to the MI position using our newly developed approach. These rearticulated mandibular models served as the experimental group. Finally, the translational, rotational, and surface deviations in the mandibular position were calculated between the experimental and control groups, and statistical analyses were performed. RESULTS: All the digital dental models were successfully articulated. Between the control and experimental groups, the largest translational difference in mandibular position was within 0.2 mm ± 0.6 mm. The largest rotational difference was within 0.1° ± 1.1°. The averaged surface deviation was 0.08 ± 0.07. The results of the Bland and Altman method of assessing measurement agreement showed tight limits for the translational, rotational, and surface deviations. In addition, the final positions of the mandibular articulated from the 80 initial positions were absolutely agreed on. CONCLUSION: The results of our study have demonstrated that using our approach, the digital dental models can be accurately and effectively articulated to the MI position. In addition, the 3-dimensional surface geometry of the mandibular teeth played a more important role in digital dental articulation than the initial position of the mandibular teeth.

The KPG index--a novel 3D classification system for maxillary canine impactions.

INTRODUCTION: Advancements in Cone Beam Computed Tomography (CBCT) have improved localization of impacted canines. The KPG index is the first 3-D classification system for classifying the position of canines based on their distance from the norm (1). The aim of this study was to determine if this index provides an estimate of the time necessary to treat an impacted canine using closed eruption. MATERIALS AND METHODS: CBCT scans of 28 impacted canines at The University of Texas School of Dentistry at Houston Department of Orthodontics were classified using the KPG index. The scores and categories were compared to the time from surgical exposure to proper positioning. RESULTS: Four canines were classified as "Easy," 11 as "Moderate," 9 as "Difficult," and 4 as "Extremely Difficult." Average treatment times associated in months were: "Easy"--11.23, "Moderate"--11.36, "Difficult"--12.76, and "Extremely Difficult"--13.23. CONCLUSIONS: The KPG index currently cannot be confirmed as an accurate means of estimating treatment time for an impacted canine. Further verification studies should include larger sample sizes and compare differing mechanics. However, there are limitations to 2-D imaging; therefore, the 3-D CBCT images and the KPG index, with further validation, will become increasingly valuable to orthodontists.

Dual Adversarial Attention Mechanism for Unsupervised Domain Adaptive Medical Image Segmentation.

Domain adaptation techniques have been demonstrated to be effective in addressing label deficiency challenges in medical image segmentation. However, conventional domain adaptation based approaches often concentrate on matching global marginal distributions between different domains in a class-agnostic fashion. In this paper, we present a dual-attention domain-adaptative segmentation network (DADASeg-Net) for cross-modality medical image segmentation. The key contribution of DADASeg-Net is a novel dual adversarial attention mechanism, which regularizes the domain adaptation module with two attention maps respectively from the space and class perspectives. Specifically, the spatial attention map guides the domain adaptation module to focus on regions that are challenging to align in adaptation. The class attention map encourages the domain adaptation module to capture class-specific instead of class-agnostic knowledge for distribution alignment. DADASeg-Net shows superior performance in two challenging medical image segmentation tasks.

Localization of Craniomaxillofacial Landmarks on CBCT Images Using 3D Mask R-CNN and Local Dependency Learning.

Cephalometric analysis relies on accurate detection of craniomaxillofacial (CMF) landmarks from cone-beam computed tomography (CBCT) images. However, due to the complexity of CMF bony structures, it is difficult to localize landmarks efficiently and accurately. In this paper, we propose a deep learning framework to tackle this challenge by jointly digitalizing 105 CMF landmarks on CBCT images. By explicitly learning the local geometrical relationships between the landmarks, our approach extends Mask R-CNN for end-to-end prediction of landmark locations. Specifically, we first apply a detection network on a down-sampled 3D image to leverage global contextual information to predict the approximate locations of the landmarks. We subsequently leverage local information provided by higher-resolution image patches to refine the landmark locations. On patients with varying non-syndromic jaw deformities, our method achieves an average detection accuracy of 1.38± 0.95mm, outperforming a related state-of-the-art method.

New 3-dimensional cephalometric analysis for orthognathic surgery.

Two basic problems have been associated with traditional 2-dimensional cephalometry. First, many important parameters cannot be measured on plain cephalograms; and second, most 2-dimensional cephalometric measurements are distorted in the presence of facial asymmetry. Three-dimensional cephalometry, which has been facilitated by the introduction of cone-beam computed tomography, can solve these problems. However, before this can be realized, fundamental problems must be solved. These include the unreliability of internal reference systems and some 3-dimensional measurements, and the lack of tools to assess and measure the symmetry. In the present report, we present a new 3-dimensional cephalometric analysis that uses different geometric approaches to solve these fundamental problems. The present analysis allows the accurate measurement of the size, shape, position, and orientation of the different facial units and incorporates a novel method to measure asymmetry.