Convolutional neural network for automated tooth segmentation on intraoral scans

BACKGROUND

Tooth segmentation on intraoral scanned (IOS) data is a prerequisite for clinical applications in digital workflows. Current state-of-the-art methods lack the robustness to handle variability in dental conditions. This study aims to propose and evaluate the performance of a convolutional neural network (CNN) model for automatic tooth segmentation on IOS images.

METHODS

A dataset of 761 IOS images (380 upper jaws, 381 lower jaws) was acquired using an intraoral scanner. The inclusion criteria included a full set of permanent teeth, teeth with orthodontic brackets, and partially edentulous dentition. A multi-step 3D U-Net pipeline was designed for automated tooth segmentation on IOS images. The model's performance was assessed in terms of time and accuracy. Additionally, the model was deployed on an online cloud-based platform, where a separate subsample of 18 IOS images was used to test the clinical applicability of the model by comparing three modes of segmentation: automated artificial intelligence-driven (A-AI), refined (R-AI), and semi-automatic (SA) segmentation.

RESULTS

The average time for automated segmentation was 31.7 ± 8.1 s per jaw. The CNN model achieved an Intersection over Union (IoU) score of 91%, with the full set of teeth achieving the highest performance and the partially edentulous group scoring the lowest. In terms of clinical applicability, SA took an average of 860.4 s per case, whereas R-AI showed a 2.6-fold decrease in time (328.5 s). Furthermore, R-AI offered higher performance and reliability compared to SA, regardless of the dentition group.

CONCLUSIONS

The 3D U-Net pipeline was accurate, efficient, and consistent for automatic tooth segmentation on IOS images. The online cloud-based platform could serve as a viable alternative for IOS segmentation.

A fine-grained orthodontics segmentation model for 3D intraoral scan data

With the widespread application of digital orthodontics in the diagnosis and treatment of oral diseases, more and more researchers focus on the accurate segmentation of teeth from intraoral scan data. The accuracy of the segmentation results will directly affect the follow-up diagnosis of dentists. Although the current research on tooth segmentation has achieved promising results, the 3D intraoral scan datasets they use are almost all indirect scans of plaster models, and only contain limited samples of abnormal teeth, so it is difficult to apply them to clinical scenarios under orthodontic treatment. The current issue is the lack of a unified and standardized dataset for analyzing and validating the effectiveness of tooth segmentation. In this work, we focus on deformed teeth segmentation and provide a fine-grained tooth segmentation dataset (3D-IOSSeg). The dataset consists of 3D intraoral scan data from more than 200 patients, with each sample labeled with a fine-grained mesh unit. Meanwhile, 3D-IOSSeg meticulously classified every tooth in the upper and lower jaws. In addition, we propose a fast graph convolutional network for 3D tooth segmentation named Fast-TGCN. In the model, the relationship between adjacent mesh cells is directly established by the naive adjacency matrix to better extract the local geometric features of the tooth. Extensive experiments show that Fast-TGCN can quickly and accurately segment teeth from the mouth with complex structures and outperforms other methods in various evaluation metrics. Moreover, we present the results of multiple classical tooth segmentation methods on this dataset, providing a comprehensive analysis of the field. All code and data will be available at https://github.com/MIVRC/Fast-TGCN.

Extraction of Tooth Cusps based on DBSCAN Density Clustering and Neighborhood Search Algorithm

A good tooth cusp extraction is helpful in evaluating the effect of cosmetic dental work in virtual tooth surgery. We propose a new tooth cusp extraction, which integrates the DBSCAN (density-based spatial clustering of applications with noise) clustering algorithm with the neighborhood search algorithm to extract tooth cusp from a three-dimensional cloud-point tooth model. This method used the point cloud height and curvature to screen out the dented point set. Then we employ the DBSCAN clustering algorithm to segment different feature regions of the tooth surface and generate the candidate point set. Finally, the candidate point set was accurately located at the tooth apex through the neighborhood search algorithm and the traversal search method of non-maximum suppression. The experimental results show that the proposed method is superior to the traditional watershed algorithm-based methods by calculating the recall rate and accuracy rate, and also has higher extraction speed and extraction precision than manual extraction methods.

Deep learning-enabled 3D multimodal fusion of cone-beam CT and intraoral mesh scans for clinically applicable tooth-bone reconstruction

High-fidelity three-dimensional (3D) models of tooth-bone structures are valuable for virtual dental treatment planning; however, they require integrating data from cone-beam computed tomography (CBCT) and intraoral scans (IOS) using methods that are either error-prone or time-consuming. Hence, this study presents Deep Dental Multimodal Fusion (DDMF), an automatic multimodal framework that reconstructs 3D tooth-bone structures using CBCT and IOS. Specifically, the DDMF framework comprises CBCT and IOS segmentation modules as well as a multimodal reconstruction module with novel pixel representation learning architectures, prior knowledge-guided losses, and geometry-based 3D fusion techniques. Experiments on real-world large-scale datasets revealed that DDMF achieved superior segmentation performance on CBCT and IOS, achieving a 0.17 mm average symmetric surface distance (ASSD) for 3D fusion with a substantial processing time reduction. Additionally, clinical applicability studies have demonstrated DDMF's potential for accurately simulating tooth-bone structures throughout the orthodontic treatment process.

A Critical Analysis of the Limitation of Deep Learning based 3D Dental Mesh Segmentation Methods in Segmenting Partial Scans

Tooth segmentation from intraoral scans is a crucial part of digital dentistry. Many Deep Learning based tooth segmentation algorithms have been developed for this task. In most of the cases, high accuracy has been achieved, although, most of the available tooth segmentation techniques make an implicit restrictive assumption of full jaw model and they report accuracy based on full jaw models. Medically, however, in certain cases, full jaw tooth scan is not required or may not be available. Given this practical issue, it is important to understand the robustness of currently available widely used Deep Learning based tooth segmentation techniques. For this purpose, we applied available segmentation techniques on partial intraoral scans and we discovered that the available deep Learning techniques under-perform drastically. The analysis and comparison presented in this work would help us in understanding the severity of the problem and allow us to develop robust tooth segmentation technique without strong assumption of full jaw model.Clinical relevance- Deep learning based tooth mesh segmentation algorithms have achieved high accuracy. In the clinical setting, robustness of deep learning based methods is of utmost importance. We discovered that the high performing tooth segmentation methods under-perform when segmenting partial intraoral scans. In our current work, we conduct extensive experiments to show the extent of this problem. We also discuss why adding partial scans to the training data of the tooth segmentation models is non-trivial. An in-depth understanding of this problem can help in developing robust tooth segmentation tenichniques.

Hierarchical Self-Supervised Learning for 3D Tooth Segmentation in Intra-Oral Mesh Scans

Accurately delineating individual teeth and the gingiva in the three-dimension (3D) intraoral scanned (IOS) mesh data plays a pivotal role in many digital dental applications, e.g., orthodontics. Recent research shows that deep learning based methods can achieve promising results for 3D tooth segmentation, however, most of them rely on high-quality labeled dataset which is usually of small scales as annotating IOS meshes requires intensive human efforts. In this paper, we propose a novel self-supervised learning framework, named STSNet, to boost the performance of 3D tooth segmentation leveraging on large-scale unlabeled IOS data. The framework follows two-stage training, i.e., pre-training and fine-tuning. In pre-training, three hierarchical-level, i.e., point-level, region-level, cross-level, contrastive losses are proposed for unsupervised representation learning on a set of predefined matched points from different augmented views. The pretrained segmentation backbone is further fine-tuned in a supervised manner with a small number of labeled IOS meshes. With the same amount of annotated samples, our method can achieve an mIoU of 89.88%, significantly outperforming the supervised counterparts. The performance gain becomes more remarkable when only a small amount of labeled samples are available. Furthermore, STSNet can achieve better performance with only 40% of the annotated samples as compared to the fully supervised baselines. To the best of our knowledge, we present the first attempt of unsupervised pre-training for 3D tooth segmentation, demonstrating its strong potential in reducing human efforts for annotation and verification.

A Tooth Segmentation Method Based on Multiple Geometric Feature Learning

Tooth segmentation is an important aspect of virtual orthodontic systems. In some existing studies using deep learning-based tooth segmentation methods, the feature learning of point coordinate information and normal vector information is not effectively distinguished. This will lead to the feature information of these two methods not producing complementary intermingling. To address this problem, a tooth segmentation method based on multiple geometric feature learning is proposed in this paper. First, the spatial transformation (T-Net) module is used to complete the alignment of dental model mesh features. Second, a multiple geometric feature learning module is designed to encode and enhance the centroid coordinates and normal vectors of each triangular mesh to highlight the differences between geometric features of different meshes. Finally, for local to global fusion features, feature downscaling and channel optimization are accomplished layer by layer using multilayer perceptron (MLP) and efficient channel attention (ECA). The experimental results show that our algorithm achieves better accuracy and efficiency of tooth segmentation and can assist dentists in their treatment work.

Two-Stream Graph Convolutional Network for Intra-Oral Scanner Image Segmentation

Precise segmentation of teeth from intra-oral scanner images is an essential task in computer-aided orthodontic surgical planning. The state-of-the-art deep learning-based methods often simply concatenate the raw geometric attributes (i.e., coordinates and normal vectors) of mesh cells to train a single-stream network for automatic intra-oral scanner image segmentation. However, since different raw attributes reveal completely different geometric information, the naive concatenation of different raw attributes at the (low-level) input stage may bring unnecessary confusion in describing and differentiating between mesh cells, thus hampering the learning of high-level geometric representations for the segmentation task. To address this issue, we design a two-stream graph convolutional network (i.e., TSGCN), which can effectively handle inter-view confusion between different raw attributes to more effectively fuse their complementary information and learn discriminative multi-view geometric representations. Specifically, our TSGCN adopts two input-specific graph-learning streams to extract complementary high-level geometric representations from coordinates and normal vectors, respectively. Then, these single-view representations are further fused by a self-attention module to adaptively balance the contributions of different views in learning more discriminative multi-view representations for accurate and fully automatic tooth segmentation. We have evaluated our TSGCN on a real-patient dataset of dental (mesh) models acquired by 3D intraoral scanners. Experimental results show that our TSGCN significantly outperforms state-of-the-art methods in 3D tooth (surface) segmentation.

Trainable WEKA (Waikato Environment for Knowledge Analysis) Segmentation Tool: Machine-Learning-Enabled Segmentation on Features of Panoramic Radiographs

Introduction: Segmentation of dental radiographs is a comprehensive subject in oral care and diagnosis. It is the process of delineating anatomical structures to simplify the diagnostic process for oral and maxillofacial radiologists.

Purpose: This paper will provide an in-depth analysis of the latest benchmarks in oral imaging by studying the segmentation of panoramic radiographs using Trainable WEKA (Waikato Environment for Knowledge Analysis) Segmentation (TWS). The aim of this research is to accurately automate segmentation where it can be implemented on a large scale of clients in order to simplify radiological diagnosis.

Methods and Materials: The experimentation was conducted by modifying open-source radiographs from UFBA UESC DENTAL IMAGES dataset. In order to simulate realistic conditions such as noise affecting regions of interest, panoramic radiographs were degraded and blurred with Gaussian noise. Accuracy was quantified by measuring the difference between the automated image and the dentist-annotated image using MorphoLibJ. To ensure the precision in results, automated predicted segmentations were observed by an oral maxillofacial radiologist and compared with the dentist-renditioning annotations of the panoramic radiographs (orthopantomograms).

Results: The TWS classifier on radiographs with an average of 32 teeth and greater (Dice value of 0.66) and an average of less than 32 teeth (F1 score of 0.59) was significant. The calculated t-value for the Jaccard index is 2.78 and the t-value for the Dice score is 2.81. The results, considering the statistical scores, were due to the independent variable. The radiographs with 32 teeth and greater had higher Intersection over Union scores and F1 scores because of less discrepancy in tooth alignment.

Conclusions: Segmentation of dental radiographs can be conducted by machine learning instead of manual segmentation.

Toward Clinically Applicable 3-Dimensional Tooth Segmentation via Deep Learning

Digital dentistry plays a pivotal role in dental health care. A critical step in many digital dental systems is to accurately delineate individual teeth and the gingiva in the 3-dimension intraoral scanned mesh data. However, previous state-of-the-art methods are either time-consuming or error prone, hence hindering their clinical applicability. This article presents an accurate, efficient, and fully automated deep learning model trained on a data set of 4,000 intraoral scanned data annotated by experienced human experts. On a holdout data set of 200 scans, our model achieves a per-face accuracy, average-area accuracy, and area under the receiver operating characteristic curve of 96.94%, 98.26%, and 0.9991, respectively, significantly outperforming the state-of-the-art baselines. In addition, our model takes only about 24 s to generate segmentation outputs, as opposed to >5 min by the baseline and 15 min by human experts. A clinical performance test of 500 patients with malocclusion and/or abnormal teeth shows that 96.9% of the segmentations are satisfactory for clinical applications, 2.9% automatically trigger alarms for human improvement, and only 0.2% of them need rework. Our research demonstrates the potential for deep learning to improve the efficacy and efficiency of dental treatment and digital dentistry.

TSegNet: An efficient and accurate tooth segmentation network on 3D dental model

Automatic and accurate segmentation of dental models is a fundamental task in computer-aided dentistry. Previous methods can achieve satisfactory segmentation results on normal dental models; however, they fail to robustly handle challenging clinical cases such as dental models with missing, crowding, or misaligned teeth before orthodontic treatments. In this paper, we propose a novel end-to-end learning-based method, called TSegNet, for robust and efficient tooth segmentation on 3D scanned point cloud data of dental models. Our algorithm detects all the teeth using a distance-aware tooth centroid voting scheme in the first stage, which ensures the accurate localization of tooth objects even with irregular positions on abnormal dental models. Then, a confidence-aware cascade segmentation module in the second stage is designed to segment each individual tooth and resolve ambiguities caused by aforementioned challenging cases. We evaluated our method on a large-scale real-world dataset consisting of dental models scanned before or after orthodontic treatments. Extensive evaluations, ablation studies and comparisons demonstrate that our method can generate accurate tooth labels robustly in various challenging cases and significantly outperforms state-of-the-art approaches by 6.5% of Dice Coefficient, 3.0% of F1 score in term of accuracy, while achieving 20 times speedup of computational time.

Deep Multi-Scale Mesh Feature Learning for Automated Labeling of Raw Dental Surfaces From 3D Intraoral Scanners

Precisely labeling teeth on digitalized 3D dental surface models is the precondition for tooth position rearrangements in orthodontic treatment planning. However, it is a challenging task primarily due to the abnormal and varying appearance of patients' teeth. The emerging utilization of intraoral scanners (IOSs) in clinics further increases the difficulty in automated tooth labeling, as the raw surfaces acquired by IOS are typically low-quality at gingival and deep intraoral regions. In recent years, some pioneering end-to-end methods (e.g., PointNet) have been proposed in the communities of computer vision and graphics to consume directly raw surface for 3D shape segmentation. Although these methods are potentially applicable to our task, most of them fail to capture fine-grained local geometric context that is critical to the identification of small teeth with varying shapes and appearances. In this paper, we propose an end-to-end deep-learning method, called MeshSegNet, for automated tooth labeling on raw dental surfaces. Using multiple raw surface attributes as inputs, MeshSegNet integrates a series of graph-constrained learning modules along its forward path to hierarchically extract multi-scale local contextual features. Then, a dense fusion strategy is applied to combine local-to-global geometric features for the learning of higher-level features for mesh cell annotation. The predictions produced by our MeshSegNet are further post-processed by a graph-cut refinement step for final segmentation. We evaluated MeshSegNet using a real-patient dataset consisting of raw maxillary surfaces acquired by 3D IOS. Experimental results, performed 5-fold cross-validation, demonstrate that MeshSegNet significantly outperforms state-of-the-art deep learning methods for 3D shape segmentation.

An automatic approach to establish clinically desired final dental occlusion for one-piece maxillary orthognathic surgery

PURPOSE

One critical step in routine orthognathic surgery is to reestablish a desired final dental occlusion. Traditionally, the final occlusion is established by hand articulating stone dental models. To date, there are still no effective solutions to establish the final occlusion in computer-aided surgical simulation. In this study, we consider the most common one-piece maxillary orthognathic surgery and propose a three-stage approach to digitally and automatically establish the desired final dental occlusion.

METHODS

The process includes three stages: (1) extraction of points of interest and teeth landmarks from a pair of upper and lower dental models; (2) establishment of Midline-Canine-Molar (M-C-M) relationship following the clinical criteria on these three regions; and (3) fine alignment of upper and lower teeth with maximum contacts without breaking the established M-C-M relationship. Our method has been quantitatively and qualitatively validated using 18 pairs of dental models.

RESULTS

Qualitatively, experienced orthodontists assess the algorithm-articulated and hand-articulated occlusions while being blind to the methods used. They agreed that occlusion results of the two methods are equally good. Quantitatively, we measure and compare the distances between selected landmarks on upper and lower teeth for both algorithm-articulated and hand-articulated occlusions. The results showed that there was no statistically significant difference between the algorithm-articulated and hand-articulated occlusions.

CONCLUSION

The proposed three-stage automatic dental articulation method is able to articulate the digital dental model to the clinically desired final occlusion accurately and efficiently. It allows doctors to completely eliminate the use of stone dental models in the future.

3D Tooth Segmentation and Labeling Using Deep Convolutional Neural Networks

In this paper, we present a novel approach for 3D dental model segmentation via deep Convolutional Neural Networks (CNNs). Traditional geometry-based methods tend to receive undesirable results due to the complex appearance of human teeth (e.g., missing/rotten teeth, feature-less regions, crowding teeth, extra medical attachments, etc.). Furthermore, labeling of individual tooth is hardly enabled in traditional tooth segmentation methods. To address these issues, we propose to learn a generic and robust segmentation model by exploiting deep Neural Networks, namely NNs. The segmentation task is achieved by labeling each mesh face. We extract a set of geometry features as face feature representations. In the training step, the network is fed with those features, and produces a probability vector, of which each element indicates the probability a face belonging to the corresponding model part. To this end, we extensively experiment with various network structures, and eventually arrive at a 2-level hierarchical CNNs structure for tooth segmentation: one for teeth-gingiva labeling and the other for inter-teeth labeling. Further, we propose a novel boundary-aware tooth simplification method to significantly improve efficiency in the stage of feature extraction. After CNNs prediction, we do graph-based label optimization and further refine the boundary with an improved version of fuzzy clustering. The accuracy of our mesh labeling method exceeds that of the state-of-art geometry-based methods, reaching 99.06 percent measured by area which is directly applicable in orthodontic CAD systems. It is also robust to any possible foreign matters on model surface, e.g., air bubbles, dental accessories, and many more.

Automatic tooth segmentation of dental mesh using a transverse plane

This paper proposes an automatic method to separate the gingiva and individual teeth from a dental mesh. We define a transverse plane that produces a cross-section of tooth lingual and labial surfaces, preserving the shape of individual teeth. The upper vertices from the transverse plane, which belong to the tooth, are projected onto the transverse plane, and partitioned into individual teeth. We apply region growing to the remaining non-segmented parts to determine the cluster the vertices belong to, and the proposed approach is fully automatic, i.e., segmentation does not require user interaction for feature point search or tooth boundary markers. The proposed segmentation method is applied to several dental mesh models to demonstrate its robustness.

Interactive Tooth Separation from Dental Model Using Segmentation Field

Tooth segmentation on dental model is an essential step of computer-aided-design systems for orthodontic virtual treatment planning. However, fast and accurate identifying cutting boundary to separate teeth from dental model still remains a challenge, due to various geometrical shapes of teeth, complex tooth arrangements, different dental model qualities, and varying degrees of crowding problems. Most segmentation approaches presented before are not able to achieve a balance between fine segmentation results and simple operating procedures with less time consumption. In this article, we present a novel, effective and efficient framework that achieves tooth segmentation based on a segmentation field, which is solved by a linear system defined by a discrete Laplace-Beltrami operator with Dirichlet boundary conditions. A set of contour lines are sampled from the smooth scalar field, and candidate cutting boundaries can be detected from concave regions with large variations of field data. The sensitivity to concave seams of the segmentation field facilitates effective tooth partition, as well as avoids obtaining appropriate curvature threshold value, which is unreliable in some case. Our tooth segmentation algorithm is robust to dental models with low quality, as well as is effective to dental models with different levels of crowding problems. The experiments, including segmentation tests of varying dental models with different complexity, experiments on dental meshes with different modeling resolutions and surface noises and comparison between our method and the morphologic skeleton segmentation method are conducted, thus demonstrating the effectiveness of our method.

Automatic Tooth Segmentation of Dental Mesh Based on Harmonic Fields

An important preprocess in computer-aided orthodontics is to segment teeth from the dental models accurately, which should involve manual interactions as few as possible. But fully automatic partition of all teeth is not a trivial task, since teeth occur in different shapes and their arrangements vary substantially from one individual to another. The difficulty is exacerbated when severe teeth malocclusion and crowding problems occur, which is a common occurrence in clinical cases. Most published methods in this area either are inaccurate or require lots of manual interactions. Motivated by the state-of-the-art general mesh segmentation methods that adopted the theory of harmonic field to detect partition boundaries, this paper proposes a novel, dental-targeted segmentation framework for dental meshes. With a specially designed weighting scheme and a strategy of a priori knowledge to guide the assignment of harmonic constraints, this method can identify teeth partition boundaries effectively. Extensive experiments and quantitative analysis demonstrate that the proposed method is able to partition high-quality teeth automatically with robustness and efficiency.

An enhanced method for registration of dental surfaces partially scanned by a 3D dental laser scanning

In this paper, we propose the fast and accurate registration method of partially scanned dental surfaces in a 3D dental laser scanning. To overcome the multiple point correspondence problems of conventional surface registration methods, we propose the novel depth map-based registration method to register 3D surface models. First, we convert a partially scanned 3D dental surface into a 2D image by generating the 2D depth map image of the surface model by applying a 3D rigid transformation into this model. Then, the image-based registration method using 2D depth map images accurately estimates the initial transformation between two consequently acquired surface models. To further increase the computational efficiency, we decompose the 3D rigid transformation into out-of-plane (i.e. x-, y-rotation, and z-translation) and in-plane (i.e. x-, y-translation, and z-rotation) transformations. For the in-plane transformation, we accelerate the transformation process by transforming the 2D depth map image instead of transforming the 3D surface model. For the more accurate registration of 3D surface models, we enhance iterative closest point (ICP) method for the subsequent fine registration. Our initial depth map-based registration well aligns each surface model. Therefore, our subsequent ICP method can accurately register two surface models since it is highly probable that the closest point pairs are the exact corresponding point pairs. The experimental results demonstrated that our method accurately registered partially scanned dental surfaces. Regarding the computational performance, our method delivered about 1.5 times faster registration than the conventional method. Our method can be successfully applied to the accurate reconstruction of 3D dental objects for orthodontic and prosthodontic treatment.

Interactive tooth partition of dental mesh base on tooth-target harmonic field

The accurate tooth partition of dental mesh is a crucial step in computer-aided orthodontics. However, tooth boundary identification is not a trivial task for tooth partition, since different shapes and their arrangements vary substantially among common clinical cases. Though curvature field is traditionally used for identifying boundaries, it is normally not reliable enough. Other methods may improve the accuracy, but require intensive user interaction. Motivated by state-of-the-art general interactive mesh segmentation methods, this paper proposes a novel tooth-target partition framework that employs harmonic fields to partition teeth accurately and effectively. In addition, a refining strategy is introduced to successfully segment teeth from the complicated dental model with indistinctive tooth boundaries on its lingual side surface, addressing an issue that had not been solved properly before. To utilise high-level information provided by the user, smart and intuitive user interfaces are also proposed with minimum interaction. In fact, most published interactive methods specifically designed for tooth partition are lacking efficient user interfaces. Extensive experiments and quantitative analyses show that our tooth partition method outperforms the state-of-the-art approaches in terms of accuracy, robustness and efficiency.