Compared with the rapidly growing artificial intelligence (AI) research in other branches of healthcare, the pace of developing AI capacities in dental care is relatively slow. Dental care automation, especially the automated capability for dental cone beam computed tomography (CBCT) segmentation and lesion detection, is highly needed. CBCT is an important imaging modality that is experiencing ever-growing utilization in various dental specialties. However, little research has been done for segmenting different structures, restorative materials, and lesions using deep learning. This is due to multifold challenges such as content-rich oral cavity and significant within-label variation on each CBCT image as well as the inherent difficulty of obtaining many high-quality labeled images for training. On the other hand, oral-anatomical knowledge exists in dentistry, which shall be leveraged and integrated into the deep learning design. In this article, we propose a novel anatomically constrained Dense U-Net for integrating oral-anatomical knowledge with data-driven Dense U-Net. The proposed algorithm is formulated as a regularized or constrained optimization and solved using mean-field variational approximation to achieve computational efficiency. Mathematical encoding for transforming descriptive knowledge into a quantitative form is also proposed. Our experiment demonstrates that the proposed algorithm outperforms the standard Dense U-Net in both lesion detection accuracy and dice coefficient (DICE) indices in multilabel segmentation. Benefited from the integration with anatomical domain knowledge, our algorithm performs well with data from a small number of patients included in the training. Note to Practitioners —This article proposes a novel deep learning algorithm to enable the automated capability for cone beam computed tomography (CBCT) segmentation and lesion detection. Despite the growing adoption of CBCT in various dental specialties, such capability is currently lacking. The proposed work will provide tools to help reduce subjectivity and human errors, as well as streamline and expedite the clinical workflow. This will greatly facilitate dental care automation. Furthermore, due to the capacity of integrating oral-anatomical knowledge into the deep learning design, the proposed algorithm does not require many high-quality labeled images to train. The algorithm can provide good accuracy under limited training samples. This ability is highly desirable for practitioners by saving labor-intensive, costly labeling efforts, and enjoying the benefits provided by AI.

Anatomically Constrained Deep Learning for Automating Dental CBCT Segmentation and Lesion Detection

Edge detection is an important task in image processing, many times as part of the segmentation process. When segmentation is performed in medical imaging, one of the preferred tools is neural networks, because of their capabilities of adaptive learning and non-linear mapping. We present in this paper the neural network tools used for edge detection and we propose one that is able to perform edge detection in dental Cone Beam Computer Tomography (CBCT) images, a necessary step for the teeth 3D reconstruction.

Neural network based edge detection for CBCT segmentation

Automatic segmentation of mandibular cortical bone is challenging due to the appearance of teeth that have
similar intensity with the bone tissue and the variety of bone intensity. In this paper we propose a new method for
automatic segmentation of mandibular cortical bone on cone-beam computed tomography (CBCT) images. The bone
tissue is segmented by using Gaussian mixture model for histogram thresholding. The mandibular inferior cortical
bone is obtained by incorporating several polynomial models to fit the structure of cortical bone on coronal slices. The
buccal and lingual cortical plate is separated by using histogram thresholding for teeth elimination and polynomial
fitting for shape extraction. After performing 3D reconstruction, the volumetric cortical bone is obtained. The proposed method gives average accuracy, sensitivity, and specificity value of 96.82%, 85.96%, 97.60%, respectively. This shows that the proposed method is promising for automatic and accurate segmentation of mandibular cortical bone on CBCT images.

/pdf/automatic-segmentation-of-mandibular-cortical-bone-on-cone-vmv7udwfdd.pdf

Automatic Segmentation of Mandibular Cortical Bone on Cone-Beam CT Images Based on Histogram Thresholding and Polynomial Fitting

Objectives:A method was proposed to segment the tooth pulp cavity region in cone beam CT) images, which aimed to make the extraction process more efficient and generate more reliable results for fu...

/pdf/cbct-image-based-segmentation-method-for-tooth-pulp-cavity-3ksh761bya.pdf

CBCT image based segmentation method for tooth pulp cavity region extraction

Cone Beam Computed Tomography (CBCT) is one of the favorite imaging technologies in dentistry because if offers 3D images in conditions of reduced irradiation. High quality 3D imaging is essential for first-rate diagnostic and treatment, despite the rather low quality images, with low contrast and high noise. The paper presents an automatic method for 3D reconstruction of the oral cavity. We analyze the existing state of the art in the field and propose a segmentation method that uses the domain knowledge related to mouth and teeth to reduce the computational load and to provide good results in an automatic procedure. The method is based on a Canny type edge detector, that was found to offer better control than active contour methods.

Automatic segmentation for 3D dental reconstruction

We will present in this paper a method for the reconstruction of the oral 3D structure using Computed Tomography (CT) sections obtained from Cone Beam Computed Tomography (CBCT). A high-quality visualization of the dental tissues, based nowadays on fine 3D reconstruction, is essential for accurate diagnosis and treatments. Dental CBCT images present several major problems because of the noisiness and of the configuration of the system. We tried several algorithms in order to overcome these issues and to obtain a feasible and efficient method that can be used directly in the clinical procedures. Our approach includes novel procedures like the delimitation of the region of interest and the combination of registration and segmentation steps, which grant faster results and open the route to better representations.

3D dental reconstruction from CBCT data

In this paper, we explore the possibilities to detect and label straightforward the teeth in Cone Beam Computed Tomography images, without performing complex segmentation procedures on teeth. We are using the Circular Hough Transform (CHT) to find the teeth positions, segmentation using intensity level to determine the mandible and deformable templates to find the best fit position for the teeth on the mandible. The dental arch is split into 16 regions and the teeth are labeled by their position in a given region.

Teeth labeling from CBCT data using the Circular Hough Transform

We present in this paper a fast method for the reconstruction of the dental 3D structure using the images obtained from Cone Beam Computed Tomography (CBCT). A high-quality visualization of the dental tissues, including 3D reconstruction, is essential for the accurate diagnosis and treatment in dentistry. The segmentation for the dental CBCT images presents several major problems because of the noisiness of the images. We analyze the existing methods and propose one that uses the information for the dental structure and distribution to reduce the computational load and to provide good results in an automatic procedure.

Andrei Ungar

Papers

Neural network based edge detection for CBCT segmentation

3D dental reconstruction from CBCT data

Teeth labeling from CBCT data using the Circular Hough Transform

Knowledge Based Segmentation for Fast 3D Dental Reconstruction from CBCT