Search

Dental caries is a bacterial infectious disease that destroys the structure of teeth. It is one of the main diseases that endanger human health [R. H. Selwitz, A. I. Ismail, and N. B. Pitts, Lancet 369(9555), 51–59 (2007)]. At present, dentists use both visual exams and radiographs for the detection of caries. Affected by the patient's dental health and the degree of caries demineralization, it is sometimes difficult to accurately identify some dental caries in x-ray images with the naked eye. Therefore, dentists need an intelligent and accurate dental caries recognition system to assist diagnosis, reduce the influence of doctors' subjective factors, and improve the efficiency of dental caries diagnosis. Therefore, this paper combines the U-Net model verified in the field of biomedical image segmentation with the convolution block attention module, designs an Attention U-Net model for caries image segmentation, and discusses the feasibility of deep learning technology in caries image recognition so as to prepare for the next clinical verification. After testing, the Dice similarity coefficient, mean pixel accuracy, mean intersection over union, and frequency-weighted intersection over the union of teeth segmentation with Attention U-Net are 95.30%, 94.46%, 93.10%, and 93.54%, respectively. The Dice similarity coefficient, mean pixel accuracy, mean intersection over union, and frequency-weighted intersection over the union of dental caries segmentation with Attention U-Net are 85.36%, 91.84%, 82.22%, and 97.08%, respectively. As a proof of concept study, this study was an initial evaluation of technology to assist dentists in the detection of caries. There is still more work needed before this can be used clinically. [ABSTRACT FROM AUTHOR]

The response rate of cancer immune checkpoint inhibitors (ICI) varies among patients, making it challenging to pre-determine whether a particular patient will respond to immunotherapy. While gene mutation is critical to the treatment outcome, a framework capable of explicitly incorporating biology knowledge has yet to be established. Here we aim to propose and validate a mutation-based deep learning model for survival analysis on 1571 patients treated with ICI. Our model achieves an average concordance index of 0.59 ± 0.13 across nine types of cancer, compared to the gold standard Cox-PH model (0.52 ± 0.10). The "black box" nature of deep learning is a major concern in healthcare field. This model's interpretability, which results from incorporating the gene pathways and protein interaction (i.e., biology-aware) rather than relying on a 'black box' approach, helps patient stratification and provides insight into novel gene biomarkers, advancing our understanding of ICI treatment. [ABSTRACT FROM AUTHOR]

A novel recursive cascaded full‐resolution residual network (RCFRR‐Net) for abdominal four‐dimensional computed tomography (4D‐CT) image registration was proposed. The entire network was end‐to‐end and trained in the unsupervised approach, which meant that the deformation vector field, which presented the ground truth, was not needed during training. The network was designed by cascading three full‐resolution residual subnetworks with different architectures. The training loss consisted of the image similarity loss and the deformation vector field regularization loss, which were calculated based on the final warped image and the fixed image, allowing all cascades to be trained jointly and perform the progressive registration cooperatively. Extensive network testing was conducted using diverse datasets, including an internal 4D‐CT dataset, a public DIRLAB 4D‐CT dataset, and a 4D cone‐beam CT (4D‐CBCT) dataset. Compared with the iteration‐based demon method and two deep learning‐based methods (VoxelMorph and recursive cascaded network), the RCFRR‐Net achieved consistent and significant gains, which demonstrated that the proposed method had superior performance and generalization capability in medical image registration. The proposed RCFRR‐Net was a promising tool for various clinical applications. [ABSTRACT FROM AUTHOR]

A nondestructive detection method that combines convolutional neural network (CNN) and photoluminescence (PL) imaging was proposed for the multi-classification and multi-grading of defects during the fabrication process of silicon solar cells. In this paper, the PL was applied to collect the images of the defects of solar cells, and an image pre-processing method was introduced for enhancing the features of the defect images. Simultaneously, the defects were defined by 13 categories and three divided grades of each under the definition rules of defects that were proposed in accordance with distribution and characteristics of each defect category, and expand data were processed by various data augmentation. The model was therefore improved and optimized based on the YOLOv5 as the feature extractor and classifier. The capability of the model on distinguishing categories and grades of solar cell defects was improved via parameter tuning and image pre-processing. Through experimental analysis, the optimal combination of hyperparameters and the actual effect of data sample pre-processing on the training results of the neural network were determined. Conclusively, the reasons for the poor recognition results of the small target defects and complex feature defects by the current model were found and further work was confirmed under the foundation of the differences in recognition results between different categories and grades. [ABSTRACT FROM AUTHOR]

This report is on convolution neural network (CNN) fusion lock-in thermography, which can implement the intelligent identification of defects for aviation honeycomb sandwich composites (HSCs). First, HSCs specimens with subsurface delamination defects were fabricated and stimulated by halogen lamps according to sinusoidal modulation, and the defects were reliably inspected using lock-in thermography. The amplitude and phase images (commonly referred to as feature images) were obtained by using a digital lock-in correlation algorithm. Furthermore, these feature images were changed into gray or color-level image formalism datasets, which is pre-processed in ways including contrast enhancement, threshold segmentation as well as mosaic data augmentation. Finally, the four-layer feature pyramid structure and ransformer are combined and introduced to the popular YOLOv5 CNN model, and a YOLOLT CNN model is formed to realize the defect identification. The average precision (AP) in the defect identification of HSCs in complex environments (contains noise and other objects) reached 93.2% and achieved an average recognition speed of 0.6 s/image. [ABSTRACT FROM AUTHOR]

The alveolar bone edge line is an important basis for the dentist to diagnose periodontitis. In view of the problem of difficulty in accurately identifying the marginal lines of alveolar bone in oral X-ray images, a periodontitis-assisted diagnosis method based deep learning is proposed. In the proposed method, the image segmentation model (PSPNet, U-Net, and Dense U-Net) was applied to segmenting oral dental X-ray images. PSPNet, U-Net, and Dense U-Net are prominent image segmentation models that have gained traction in the field of medical X-ray image segmentation. These models have demonstrated their effectiveness in accurately segmenting regions of interest in medical images, and have been extensively employed in various medical imaging segmentations. Based on this, the Canny algorithm was used to extract the alveolar bone edge line from the segmentation image. According to the experiment results, the Dense U-Net achieves the best oral dental X-ray image segmentation results. The Dice, Jc, HD95, and ASD of Dense U-Net for the segmentation of 140 periodontitis oral dental X-ray images in the test set are 95.21, 90.85, 34.53, and 34.67%, respectively. Compared with PSPNet and U-Net, the oral dental X-ray image segmentation results of the Dense U-Net model have significant advantages in four metrics: Dice, Jc, HD95, and ASD. Moreover, the alveolar bone edge lines extracted by the Canny algorithm from the oral dental X-ray images segmented by Dense U-Net are closer to the manually labeled results. Experimental results fully demonstrate the effectiveness of using deep learning techniques to extract the alveolar bone edge line. This study has provided a useful exploration in assisting dentists to better diagnose periodontitis. [ABSTRACT FROM AUTHOR]

Wood defects are quickly identified from an optical image based on deep learning methodology, which effectively improves the wood utilization. The traditional neural network technique is unemployed for the wood defect detection of optical image used, which results from a long training time, low recognition accuracy, and nonautomatic extraction of defect image features. In this paper, a wood knot defect detection model (so-called BLNN) combined deep learning is reported. Two subnetworks composed of convolutional neural networks are trained by Pytorch. By using the feature extraction capabilities of the two subnetworks and combining the bilinear join operation, the fine-grained features of the image are obtained. The experimental results show that the accuracy has reached up 99.20%, and the training time is obviously reduced with the speed of defect detection about 0.0795 s/image. It indicates that BLNN has the ability to improve the accuracy of defect recognition and has a potential application in the detection of wood knot defects. [ABSTRACT FROM AUTHOR]

Wood defects are quickly identified from an optical image based on deep learning methodology, which effectively improves wood utilization. Traditional neural network techniques have not yet been employed for wood defect detection due to long training time, low recognition accuracy, and nonautomatical extraction of defect image features. In this work, a model (so-called ReSENet-18) for wood knot defect detection that combined deep learning and transfer learning is proposed. The "squeeze-and-excitation" (SE) module is firstly embedded into the "residual basic block" structure for a "SE-Basic-Block" module construction. This model has the advantages of the features that are extracted in the channel dimension, and it is fused in multiscale with original features. Instantaneously, the fully connected layer is replaced with a global average pooling; consequently, the model parameters could be reduced effectively. The experimental results show that the accuracy has reached 99.02%, meanwhile the training time is also reduced. It shows that the proposed deep convolutional neural network based on ReSENet-18 combined with transfer learning can improve the accuracy of defect recognition and has a potential application in the detection of wood knot defects. [ABSTRACT FROM AUTHOR]

• This paper proposes an innovative approach that integrates neural networks with photoluminescence detection technology to address defects such as cracks, dirt, dark spots, and scratches in solar cells. • The YOLOv5 model undergoes optimization in three distinct stages, encompassing global optimization, neck network structure refinement, and head structure enhancement. These optimizations collectively elevate the model's capabilities in recognizing, detecting, and classifying defects in solar cells. • Advanced techniques are employed, including the introduction of structures such as Transformer Encoder, the incorporation of additional small target detection boxes, utilization of Bi-FPN structure, and insertion of CBAM structure blocks. These optimizations contribute to the improvement of the model's feature recognition and extraction capabilities, further enhancing its ability to identify defects. In the production process of solar cells, inevitable faults such as cracks, dirt, dark spots, and scratches may occur, which could potentially impact the lifespan and power generation efficiency of solar cells. Addressing this issue, this paper combines neural networks with photoluminescence detection technology and proposes a novel neural network model for the classification and grading of defects in solar cells. Firstly, the YOLOv5 model is optimized and adjusted for algorithm and network structure. The optimization process is divided into three parts: global optimization of the network structure, optimization of the neck network structure, and optimization of the head structure, each addressing specific issues in recognition, detection, and classification. The impact of the optimized network model on recognition and detection speed is analyzed, and solutions are proposed to address any observed effects. Additionally, an iterative update of neural network hyperparameter combinations is performed for solar cell defect identification. Finally, using the ultimately optimized model structure in conjunction with the optimal hyperparameter combination, comparative experiments are conducted on neural networks for different target identification using the photoluminescence characteristics dataset of solar cells. The recognition improvement of the optimized model and its differences from other models are analyzed. [ABSTRACT FROM AUTHOR]

• A new non-destructive detection method that integrates a convolutional neural network and high-power halogen lamp-induced thermal-wave diffusion multidimensional features were introduced to realize the intelligent classification and quantitative characterization of honeycomb sandwich composites (HSCs) defects. • YOLOLIT is designed to achieve intelligent classification and identification of defects and quantitative characterization of defect sizes. • The optimal hyperparameter combination is obtained by iterating the hyperparameters of the neural network using a genetic algorithm. HSCs defects may have a serious impact on the mechanical properties of the material. Accurate and reliable non-destructive defect detection of materials can reduce safety hazards. Therefore, a reliable technology for HSCs defect detection is needed. In this present study, a new non-destructive detection method that integrates a convolutional neural network (CNN) and high-power halogen lamp-induced thermal-wave diffusion multidimensional features were introduced to realize the intelligent classification and quantitative characterization of honeycomb sandwich composites (HSCs) defects. Firstly, HSC specimens with delamination defects were detected by the halogen lamp-induced lock-in thermography (LIT) algorithm, and feature images were extracted. After image pre-processing, the images are used as defect feature datasets. Furthermore, combining the four-layer feature pyramid structure and Transformer, a CNN model named YOLOLIT is designed to achieve intelligent classification and identification of defects and quantitative characterization of defect sizes. Subsequently, the optimal hyperparameter combination is obtained by iterating the hyperparameters of the neural network using a genetic algorithm. In a complex environment, accurate recognition of circular defects with a minimum diameter of Φ6mm is achieved, and the mAP of up defect is 90.2%, down defect is 86.2%, remove defect is 72.9%, and the average recognition speed of 0.2 s/image is achieved. Finally, the relation between defect size and detection error as well as that between LIT frequency and detection error is studied. The detection error range and optimal LIT frequency are obtained. The experimental results illustrated this method may be a potential infrared non-destructive defect detection solution. [ABSTRACT FROM AUTHOR]