Attention-guided jaw bone lesion diagnosis in panoramic radiography using minimal labeling effort.

Developing a deep-learning-based diagnostic model demands extensive labor for medical image labeling. Attempts to reduce the labor often lead to incomplete or inaccurate labeling, limiting the diagnostic performance of models. This paper (i) constructs an attention-guiding framework that enhances the diagnostic performance of jaw bone pathology by utilizing attention information with partially labeled data; (ii) introduces an additional loss to minimize the discrepancy between network attention and its label; (iii) introduces a trapezoid augmentation method to maximize the utility of minimally labeled data. The dataset includes 716 panoramic radiograph data for jaw bone lesions and normal cases collected and labeled by two radiologists from January 2019 to February 2021. Experiments show that guiding network attention with even 5% of attention-labeled data can enhance the diagnostic accuracy for pathology from 92.41 to 96.57%. Furthermore, ablation studies reveal that the proposed augmentation methods outperform prior preprocessing and augmentation combinations, achieving an accuracy of 99.17%. The results affirm the capability of the proposed framework in fine-grained diagnosis using minimally labeled data, offering a practical solution to the challenges of medical image analysis.

Resolving the non-uniformity in the feature space of age estimation: A deep learning model based on feature clusters of panoramic images.

Age estimation is important in forensics, and numerous techniques have been investigated to estimate age based on various parts of the body. Among them, dental tissue is considered reliable for estimating age as it is less influenced by external factors. The advancement in deep learning has led to the development of automatic estimation of age using dental panoramic images. Typically, most of the medical datasets used for model learning are non-uniform in the feature space. This causes the model to be highly influenced by dense feature areas, resulting in adequate estimations; however, relatively poor estimations are observed in other areas. An effective solution to address this issue can be pre-dividing the data by age feature and training each regressor to estimate the age for individual features. In this study, we divide the data based on feature clusters obtained from unsupervised learning. The developed model comprises a classification head and multi-regression head, wherein the former predicts the cluster to which the data belong and the latter estimates the age within the predicted cluster. The visualization results show that the model can focus on a clinically meaningful area in each cluster for estimating age. The proposed model outperforms the models without feature clusters by focusing on the differences within the area. The performance improvement is particularly noticeable in the growth and aging periods. Furthermore, the model can adequately estimate the age even for samples with a high probability of classification error as they are located at the border of two feature clusters.

Estimating chronological age through learning local and global features of panoramic radiographs in the Korean population.

This study suggests a hybrid method based on ResNet50 and vision transformer (ViT) in an age estimation model. To this end, panoramic radiographs are used for learning by considering both local features and global information, which is important in estimating age. Transverse and longitudinal panoramic images of 9663 patients were selected (4774 males and 4889 females with a mean age of 39 years and 3 months). To compare ResNet50, ViT, and the hybrid model, the mean absolute error, mean square error, root mean square error, and coefficient of determination (R2) were used as metrics. The results confirmed that the age estimation model designed using the hybrid method performed better than those using only ResNet50 or ViT. The estimation is highly accurate for young people at an age with distinct growth characteristics. When examining the basis for age estimation in the hybrid model through attention rollout, the proposed model used logical and important factors rather than relying on unclear elements as the basis for age estimation.

Deep learning synthesis of cone-beam computed tomography from zero echo time magnetic resonance imaging.

Cone-beam computed tomography (CBCT) produces high-resolution of hard tissue even in small voxel size, but the process is associated with radiation exposure and poor soft tissue imaging. Thus, we synthesized a CBCT image from the magnetic resonance imaging (MRI), using deep learning and to assess its clinical accuracy. We collected patients who underwent both CBCT and MRI simultaneously in our institution (Seoul). MRI data were registered with CBCT data, and both data were prepared into 512 slices of axial, sagittal, and coronal sections. A deep learning-based synthesis model was trained and the output data were evaluated by comparing the original and synthetic CBCT (syCBCT). According to expert evaluation, syCBCT images showed better performance in terms of artifacts and noise criteria but had poor resolution compared to the original CBCT images. In syCBCT, hard tissue showed better clarity with significantly different MAE and SSIM. This study result would be a basis for replacing CBCT with non-radiation imaging that would be helpful for patients planning to undergo both MRI and CBCT.

Three-Dimensional Postoperative Results Prediction for Orthognathic Surgery through Deep Learning-Based Alignment Network.

To date, for the diagnosis of dentofacial dysmorphosis, we have relied almost entirely on reference points, planes, and angles. This is time consuming, and it is also greatly influenced by the skill level of the practitioner. To solve this problem, we wanted to know if deep neural networks could predict postoperative results of orthognathic surgery without relying on reference points, planes, and angles. We use three-dimensional point cloud data of the skull of 269 patients. The proposed method has two main stages for prediction. In step 1, the skull is divided into six parts through the segmentation network. In step 2, three-dimensional transformation parameters are predicted through the alignment network. The ground truth values of transformation parameters are calculated through the iterative closest points (ICP), which align the preoperative part of skull to the corresponding postoperative part of skull. We compare pointnet, pointnet++ and pointconv for the feature extractor of the alignment network. Moreover, we design a new loss function, which considers the distance error of transformed points for a better accuracy. The accuracy, mean intersection over union (mIoU), and dice coefficient (DC) of the first segmentation network, which divides the upper and lower part of skull, are 0.9998, 0.9994, and 0.9998, respectively. For the second segmentation network, which divides the lower part of skull into 5 parts, they were 0.9949, 0.9900, 0.9949, respectively. The mean absolute error of transverse, anterior-posterior, and vertical distance of part 2 (maxilla) are 0.765 mm, 1.455 mm, and 1.392 mm, respectively. For part 3 (mandible), they were 1.069 mm, 1.831 mm, and 1.375 mm, respectively, and for part 4 (chin), they were 1.913 mm, 2.340 mm, and 1.257 mm, respectively. From this study, postoperative results can now be easily predicted by simply entering the point cloud data of computed tomography.

Deep Learning-Based Prediction of Paresthesia after Third Molar Extraction: A Preliminary Study.

The purpose of this study was to determine whether convolutional neural networks (CNNs) can predict paresthesia of the inferior alveolar nerve using panoramic radiographic images before extraction of the mandibular third molar. The dataset consisted of a total of 300 preoperative panoramic radiographic images of patients who had planned mandibular third molar extraction. A total of 100 images taken of patients who had paresthesia after tooth extraction were classified as Group 1, and 200 images taken of patients without paresthesia were classified as Group 2. The dataset was randomly divided into a training and validation set (n = 150 [50%]), and a test set (n = 150 [50%]). CNNs of SSD300 and ResNet-18 were used for deep learning. The average accuracy, sensitivity, specificity, and area under the curve were 0.827, 0.84, 0.82, and 0.917, respectively. This study revealed that CNNs can assist in the prediction of paresthesia of the inferior alveolar nerve after third molar extraction using panoramic radiographic images.

Deep learning neural networks to differentiate Stafne's bone cavity from pathological radiolucent lesions of the mandible in heterogeneous panoramic radiography.

This study aimed to develop a high-performance deep learning algorithm to differentiate Stafne's bone cavity (SBC) from cysts and tumors of the jaw based on images acquired from various panoramic radiographic systems. Data sets included 176 Stafne's bone cavities and 282 odontogenic cysts and tumors of the mandible (98 dentigerous cysts, 91 odontogenic keratocysts, and 93 ameloblastomas) that required surgical removal. Panoramic radiographs were obtained using three different imaging systems. The trained model showed 99.25% accuracy, 98.08% sensitivity, and 100% specificity for SBC classification and resulted in one misclassified SBC case. The algorithm was approved to recognize the typical imaging features of SBC in panoramic radiography regardless of the imaging system when traced back with Grad-Cam and Guided Grad-Cam methods. The deep learning model for SBC differentiating from odontogenic cysts and tumors showed high performance with images obtained from multiple panoramic systems. The present algorithm is expected to be a useful tool for clinicians, as it diagnoses SBCs in panoramic radiography to prevent unnecessary examinations for patients. Additionally, it would provide support for clinicians to determine further examinations or referrals to surgeons for cases where even experts are unsure of diagnosis using panoramic radiography alone.

Linearly Sensitive Pressure Sensor Based on a Porous Multistacked Composite Structure with Controlled Mechanical and Electrical Properties.

Capacitive pressure sensors based on porous structures have been extensively explored for various applications because their sensing performance is superior to that of conventional polymer sensors. However, it is challenging to develop sufficiently sensitive pressure sensors with linearity over a wide pressure range owing to the trade-off between linearity and sensitivity. This study demonstrates a novel strategy for the fabrication of a pressure sensor consisting of stacked carbon nanotubes (CNTs) and polydimethylsiloxane. With the addition of carbon nanotubes, the structure is linearly compressed due to the reinforced mechanical properties, thereby resulting in high linearity. Additionally, the percolation effect is boosted by the CNTs having a high dielectric constant, thus improving the sensitivity. The pressure sensor exhibits linear sensitivity (R2 = 0.991) in the medium-pressure range (10-100 kPa). Furthermore, it delivers excellent performance with a fast response time (∼60 ms), in conjunction with high repeatability, reproducibility, and reliability (5 and 50 kPa/1000 cycles). The fabricated sensors are applied in wearable devices to monitor finger bending and detect finger motions in real time with high precision. The large-area sensor is integrated with a neural network to accurately recognize the sitting posture on a plane, thereby demonstrating the wide-range detection performance.

Deep-Learning-Based Detection of Cranio-Spinal Differences between Skeletal Classification Using Cephalometric Radiography.

The aim of this study was to reveal cranio-spinal differences between skeletal classification using convolutional neural networks (CNNs). Transverse and longitudinal cephalometric images of 832 patients were used for training and testing of CNNs (365 males and 467 females). Labeling was performed such that the jawbone was sufficiently masked, while the parts other than the jawbone were minimally masked. DenseNet was used as the feature extractor. Five random sampling crossvalidations were performed for two datasets. The average and maximum accuracy of the five crossvalidations were 90.43% and 92.54% for test 1 (evaluation of the entire posterior-anterior (PA) and lateral cephalometric images) and 88.17% and 88.70% for test 2 (evaluation of the PA and lateral cephalometric images obscuring the mandible). In this study, we found that even when jawbones of class I (normal mandible), class II (retrognathism), and class III (prognathism) are masked, their identification is possible through deep learning applied only in the cranio-spinal area. This suggests that cranio-spinal differences between each class exist.

Deep learning based prediction of necessity for orthognathic surgery of skeletal malocclusion using cephalogram in Korean individuals.

BACKGROUND: Posteroanterior and lateral cephalogram have been widely used for evaluating the necessity of orthognathic surgery. The purpose of this study was to develop a deep learning network to automatically predict the need for orthodontic surgery using cephalogram. METHODS: The cephalograms of 840 patients (Class ll: 244, Class lll: 447, Facial asymmetry: 149) complaining about dentofacial dysmorphosis and/or a malocclusion were included. Patients who did not require orthognathic surgery were classified as Group I (622 patients-Class ll: 221, Class lll: 312, Facial asymmetry: 89). Group II (218 patients-Class ll: 23, Class lll: 135, Facial asymmetry: 60) was set for cases requiring surgery. A dataset was extracted using random sampling and was composed of training, validation, and test sets. The ratio of the sets was 4:1:5. PyTorch was used as the framework for the experiment. RESULTS: Subsequently, 394 out of a total of 413 test data were properly classified. The accuracy, sensitivity, and specificity were 0.954, 0.844, and 0.993, respectively. CONCLUSION: It was found that a convolutional neural network can determine the need for orthognathic surgery with relative accuracy when using cephalogram.

Deep learning based prediction of extraction difficulty for mandibular third molars.

This paper proposes a convolutional neural network (CNN)-based deep learning model for predicting the difficulty of extracting a mandibular third molar using a panoramic radiographic image. The applied dataset includes a total of 1053 mandibular third molars from 600 preoperative panoramic radiographic images. The extraction difficulty was evaluated based on the consensus of three human observers using the Pederson difficulty score (PDS). The classification model used a ResNet-34 pretrained on the ImageNet dataset. The correlation between the PDS values determined by the proposed model and those measured by the experts was calculated. The prediction accuracies for C1 (depth), C2 (ramal relationship), and C3 (angulation) were 78.91%, 82.03%, and 90.23%, respectively. The results confirm that the proposed CNN-based deep learning model could be used to predict the difficulty of extracting a mandibular third molar using a panoramic radiographic image.

Deep-learning for predicting C-shaped canals in mandibular second molars on panoramic radiographs.

OBJECTIVE: The aim of this study was to evaluate the use of a convolutional neural network (CNN) system for predicting C-shaped canals in mandibular second molars on panoramic radiographs. METHODS: Panoramic and cone beam CT (CBCT) images obtained from June 2018 to May 2020 were screened and 1020 patients were selected. Our dataset of 2040 sound mandibular second molars comprised 887 C-shaped canals and 1153 non-C-shaped canals. To confirm the presence of a C-shaped canal, CBCT images were analyzed by a radiologist and set as the gold standard. A CNN-based deep-learning model for predicting C-shaped canals was built using Xception. The training and test sets were set to 80 to 20%, respectively. Diagnostic performance was evaluated using accuracy, sensitivity, specificity, and precision. Receiver-operating characteristics (ROC) curves were drawn, and the area under the curve (AUC) values were calculated. Further, gradient-weighted class activation maps (Grad-CAM) were generated to localize the anatomy that contributed to the predictions. RESULTS: The accuracy, sensitivity, specificity, and precision of the CNN model were 95.1, 92.7, 97.0, and 95.9%, respectively. Grad-CAM analysis showed that the CNN model mainly identified root canal shapes converging into the apex to predict the C-shaped canals, while the root furcation was predominantly used for predicting the non-C-shaped canals. CONCLUSIONS: The deep-learning system had significant accuracy in predicting C-shaped canals of mandibular second molars on panoramic radiographs.

Deep learning based discrimination of soft tissue profiles requiring orthognathic surgery by facial photographs.

Facial photographs of the subjects are often used in the diagnosis process of orthognathic surgery. The aim of this study was to determine whether convolutional neural networks (CNNs) can judge soft tissue profiles requiring orthognathic surgery using facial photographs alone. 822 subjects with dentofacial dysmorphosis and / or malocclusion were included. Facial photographs of front and right side were taken from all patients. Subjects who did not need orthognathic surgery were classified as Group I (411 subjects). Group II (411 subjects) was set up for cases requiring surgery. CNNs of VGG19 was used for machine learning. 366 of the total 410 data were correctly classified, yielding 89.3% accuracy. The values of accuracy, precision, recall, and F1 scores were 0.893, 0.912, 0.867, and 0.889, respectively. As a result of this study, it was found that CNNs can judge soft tissue profiles requiring orthognathic surgery relatively accurately with the photographs alone.

Room-Temperature Hydrogen-Gas Sensor Based on Carbon Nanotube Yarn.

The proposed study describes the development of a carbon nanotube (CNT)-based gas sensor capable of detecting the presence of hydrogen (H2) gas at room temperature. CNT yarn used in the proposed sensor was fabricated from synthesized CNT arrays. Subsequently, the yarn was treated by means of a simple one-step procedure, called acid treatment, to facilitate removal of impurities from the yarn surface and forming functional species. To verify the proposed sensor's effectiveness with regard to detection of H2 gas at room temperature, acid-treated CNT and pure yarns were fabricated and tested under identical conditions. Corresponding results demonstrate that compared to the untreated CNT yarn, the acid-treated CNT yarn exhibits higher sensitivity to the presence of H2 gas at room temperature. Additionally, the acid-treated CNT yarn was observed to demonstrate excellent selectivity pertaining to H2 gas.

H2 Gas Sensor Based on Pd-Loaded Carbon Nanotube Film.

Palladium-coated multi-walled carbon nanotube (Pd-MWCNT) nanocomposites have been experimentally proven to show highly improved hydrogen (H2) gas detection characteristics at room temperature when compared with single MWCNTs. In this context, we develop an efficient and convenient method for forming nanocomposites by coating Pd nanoparticles on an MWCNT film. Furthermore, we test the applicability of the nanocomposites as sensing materials in detecting H2 gas at room temperature in a reliable and sensitive manner in contrast with ordinary metal-oxidebased gas sensors that operate at high temperatures. We first study the detection efficacy of the Pd-MWCNT film relative to pure MWCNT film. Subsequently, we investigate the Pd-MWCNT sensor's sensitivity over time for different gas concentrations, the sensor response time, and sensor reproducibility and reliability under various conditions including bending tests. Our sensor exhibits stable reliable detection characteristics and excellent structural flexibility.

Steel-surface defect detection using a switching-lighting scheme.

In this paper a novel filtering scheme combined with a lighting method is proposed for defect detection in steel surfaces. A steel surface has non-uniform brightness and various shaped defects, which cause difficulties in defect detection. To solve this problem we propose a sub-optimal filtering that is combined with a switching-lighting method. First, dual-light switching lighting (DLSL) is explained, which decreases the effect of non-uniformity of surface brightness and improves the detection accuracy. By using the DLSL method, defects are represented as alternated black and white patterns regardless of the size, shape, or orientation of defects. Therefore, defects can be detected by finding alternated black and white patterns. Second, we propose a scheme for detecting defects in steel-surface images acquired using the DLSL method. The presence of scales strongly affects the optical properties of the surface. Moreover, the textures of steel-plate images vary greatly because of the temperature and grade of steel. Therefore, conventional filter-design methods are not effective for different image textures. A sub-optimal scheme based on an optimized general-finite impulse-response filter is also proposed. Finally, experimental results conducted on steel-surface images from an actual steel-production line show the effectiveness of the proposed algorithm.

Algorithm for detecting seam cracks in steel plates using a Gabor filter combination method.

Presently, product inspection based on vision systems is an important part of the steel-manufacturing industry. In this work, we focus on the detection of seam cracks in the edge region of steel plates. Seam cracks are generated in the vertical direction, and their width range is 0.2-0.6 mm. Moreover, the gray values of seam cracks are only 20-30 gray levels lower than those of the neighboring surface. Owing to these characteristics, we propose a new algorithm for detecting seam cracks using a Gabor filter combination method. To enhance the performance, we extracted features of seam cracks and employed a support vector machine classifier. The experimental results show that the proposed algorithm is suitable for detecting seam cracks.

Defect detection for corner cracks in steel billets using a wavelet reconstruction method.

Presently, automatic inspection algorithms are widely used to ensure high-quality products and achieve high productivity in the steelmaking industry. In this paper, we propose a vision-based method for detecting corner cracks on the surface of steel billets. Because of the presence of scales composed of oxidized substances, the billet surfaces are not uniform and vary considerably with the lighting conditions. To minimize the influence of scales and improve the accuracy of detection, a detection method based on a visual inspection algorithm is proposed. Wavelet reconstruction is used to reduce the effect of scales. Texture and morphological features are used to identify the corner cracks among the defective candidates. Finally, the experimental results show that the proposed algorithm is effective in detecting corner cracks on the surfaces of the steel billets.

Real-time defect detection of steel wire rods using wavelet filters optimized by univariate dynamic encoding algorithm for searches.

We propose a new defect detection algorithm for scale-covered steel wire rods. The algorithm incorporates an adaptive wavelet filter that is designed on the basis of lattice parameterization of orthogonal wavelet bases. This approach offers the opportunity to design orthogonal wavelet filters via optimization methods. To improve the performance and the flexibility of wavelet design, we propose the use of the undecimated discrete wavelet transform, and separate design of column and row wavelet filters but with a common cost function. The coefficients of the wavelet filters are optimized by the so-called univariate dynamic encoding algorithm for searches (uDEAS), which searches the minimum value of a cost function designed to maximize the energy difference between defects and background noise. Moreover, for improved detection accuracy, we propose an enhanced double-threshold method. Experimental results for steel wire rod surface images obtained from actual steel production lines show that the proposed algorithm is effective.

Pinhole detection in steel slab images using Gabor filter and morphological features.

Presently, product inspection for quality control is becoming an important part in the steel manufacturing industry. In this paper, we propose a vision-based method for detection of pinholes in the surface of scarfed slabs. The pinhole is a very tiny defect that is 1-5 mm in diameter. Because the brightness in the surface of a scarfed slab is not uniform and the size of a pinhole is small, it is difficult to detect pinholes. To overcome the above-mentioned difficulties, we propose a new defect detection algorithm using a Gabor filter and morphological features. The Gabor filter was used to extract defective candidates. The morphological features are used to identify the pinholes among the defective candidates. Finally, the experimental results show that the proposed algorithm is effective to detect pinholes in the surface of the scarfed slab.