Objectives: The purpose of this study was to automatically diagnose odontogenic cysts and tumors of both jaws on panoramic radiographs using deep learning. We proposed a novel framework of deep convolution neural network (CNN) with data augmentation for detection and classification of the multiple diseases. Methods: We developed a deep CNN modified from YOLOv3 for detecting and classifying odontogenic cysts and tumors of both jaws. Our data set of 1282 panoramic radiographs comprised 350 dentigerous cysts (DCs), 302 periapical cysts (PCs), 300 odontogenic keratocysts (OKCs), 230 ameloblastomas (ABs), and 100 normal jaws with no disease. In addition, the number of radiographs was augmented 12-fold by flip, rotation, and intensity changes. We evaluated the classification performance of the developed CNN by calculating sensitivity, specificity, accuracy, and area under the curve (AUC) for diseases of both jaws. Results: The overall classification performance for the diseases improved from 78.2% sensitivity, 93.9% specificity,91.3% accuracy, and 0.86 AUC using the CNN with unaugmented data set to 88.9% sensitivity, 97.2% specificity, 95.6% accuracy, and 0.94 AUC using the CNN with augmented data set. CNN using augmented data set had the following sensitivities, specificities, accuracies, and AUCs: 91.4%, 99.2%, 97.8%, and 0.96 for DCs, 82.8%, 99.2%, 96.2%, and 0.92 for PCs, 98.4%,92.3%,94.0%, and 0.97 for OKCs, 71.7%, 100%, 94.3%, and 0.86 for ABs, and 100.0%, 95.1%, 96.0%, and 0.97 for normal jaws, respectively. Conclusion: The CNN method we developed for automatically diagnosing odontogenic cysts and tumors of both jaws on panoramic radiographs using data augmentation showed high sensitivity, specificity, accuracy, and AUC despite the limited number of panoramic images involved.

Song IS, et al. Imaging Sci Dent. 2022 Jul;52:e37. https://doi.org/10.5624/isd.20220078

Disaster victim identification issues are especially critical and urgent after a large-scale disaster. The aim of this study was to suggest an automatic detection of natural teeth and dental treatment patterns based on dental panoramic radiographs (DPRs) using deep learning to promote its applicability as human identifiers. A total of 1 638 DPRs, of which the chronological age ranged from 20 to 49 years old, were collected from January 2000 to November 2020. This dataset consisted of natural teeth, prostheses, teeth with root canal treatment, and implants. The detection of natural teeth and dental treatment patterns including the identification of teeth number was done with a pre-trained object detection network which was a convolutional neural network modified by EfficientDet-D3. The objective metrics for the average precision were 99.1% for natural teeth, 80.6% for prostheses, 81.2% for treated root canals, and 96.8% for implants, respectively. The values for the average recall were 99.6%, 84.3%, 89.2%, and 98.1%, in the same order, respectively. This study showed outstanding performance of convolutional neural network using dental panoramic radiographs in automatically identifying teeth number and detecting natural teeth, prostheses, treated root canals, and implants. Key PointsIt is useful to use dental panoramic radiographs to perform the disaster victim identification (DVI).Individual dental treatment patterns could be unique and powerful characteristics as human identifier.Deep learning can automatically detect teeth and dental treatment patterns with favourable accuracy.Constructing a big database including detected teeth information would make it possible to fully automate entire process of DVI.

The purpose of this study was to develop a complete digital workflow for planning, simulation, and evaluation for orthognathic surgery based on 3D digital natural head position reproduction, a cloud-based collaboration platform, and 3D landmark-based evaluation. We included 24 patients who underwent bimaxillary orthognathic surgery. Surgeons and engineers could share the massive image data immediately and conveniently and collaborate closely in surgical planning and simulation using a cloud-based platform. The digital surgical splint could be optimized for a specific patient before or after the physical fabrication of 3D printing splints through close collaboration. The surgical accuracy was evaluated comprehensively via the translational (linear) and rotational (angular) discrepancies between identical 3D landmarks on the simulation and postoperative computed tomography (CT) models. The means of the absolute linear discrepancy at eight tooth landmarks were 0.61 ± 0.55, 0.86 ± 0.68, and 1.00 ± 0.79 mm in left–right, advance–setback, and impaction–elongation directions, respectively, and 1.67 mm in the root mean square direction. The linear discrepancy in the left–right direction was significantly different from the other two directions as shown by analysis of variance (ANOVA, p < 0.05). The means of the absolute angular discrepancies were 1.43 ± 1.06°, 0.50 ± 0.31°, and 0.58 ± 0.41° in the pitch, roll, and yaw orientations, respectively. The angular discrepancy in the pitch orientation was significantly different from the other two orientations (ANOVA, p < 0.05). The complete digital workflow that we developed for orthognathic patients provides efficient and streamlined procedures for orthognathic surgery and shows high surgical accuracy with efficient image data sharing and close collaboration.

Cone-beam CT (CBCT) is widely used in dental clinics but exhibits limitations in assessing soft tissue pathology because of its lack of contrast resolution and low Hounsfield Units (HU) quantification accuracy. We aimed to increase the image quality and HU accuracy of CBCTs while preserving anatomical structures. We generated CT-like images from CBCT images using a patchwise contrastive learning-based GAN model. Our model was trained on unpaired CT and CBCT datasets with the novel combination of losses and the feature extractor pretrained on our training dataset. We evaluated the quality of the images generated by our model in terms of Fréchet inception distance (FID), peak signal-to-noise ratio (PSNR), mean absolute error (MAE), and root mean square error (RMSE). Additionally, the structure preservation performance was assessed by the structure score. As a result, the generated CT-like images by our model were significantly superior to those generated by various baseline models in terms of FID, PSNR, MAE, RMSE, and structure score. Therefore, we demonstrated that our model provided the complementary benefits of preserving the anatomical structures of the input CBCT images and improving the image quality to be similar to those of CT images.