Content‐based gastric image retrieval using convolutional neural networks

The endoscopy procedure has demonstrated great efficiency in detecting stomach lesions, with extensive numbers of endoscope images produced globally each day. The content‐based gastric image retrieval (CBGIR) system has demonstrated substantial potential in gastric image analysis. Gastric precancerous diseases (GPD) have higher prevalence in gastric cancer patients. Thus, effective intervention is crucial at the GPD stage. In this paper, a CBGIR method is proposed using a modified ResNet‐18 to generate binary hash codes for a rapid and accurate image retrieval process. We tested several popular models (AlexNet, VGGNet and ResNet), with ResNet‐18 determined as the optimum option. Our proposed method was valued using a GPD data set, resulting in a classification accuracy of 96.21 ± 0.66% and a mean average precision of 0.927 ± 0.006 , outperforming other state‐of‐art conventional methods. Furthermore, we constructed a Gastric‐Map (GM) based on feature representations in order to visualize the retrieval results. This work has great auxiliary significance for endoscopists in terms of understanding the typical GPD characteristics and improving aided diagnosis.     

Machine and component residual life estimation through the application of neural networks

This paper concerns the use of neural networks for predicting the residual life of machines and components. In addition, the advantage of using condition-monitoring data to enhance the predictive capability of these neural networks was also investigated. A number of neural network variations were trained and tested with the data of two different reliability-related datasets. The first dataset represents the renewal case where the failed unit is repaired and restored to a good-as-new condition. Data were collected in the laboratory by subjecting a series of similar test pieces to fatigue loading with a hydraulic actuator. The average prediction error of the various neural networks being compared varied from 431 to 841 s on this dataset, where test pieces had a characteristic life of 8971 s. The second dataset were collected from a group of pumps used to circulate a water and magnetite solution within a plant. The data therefore originated from a repaired system affected by reliability degradation. When optimized, the multi-layer perceptron neural networks trained with the Levenberg-Marquardt algorithm and the general regression neural network produced a sum-of-squares error within 11.1% of each other for the renewal dataset. The small number of inputs and poorly mapped input space on the second dataset meant that much larger errors were recorded on some of the test data. The potential for using neural networks for residual life prediction and the advantage of incorporating condition-based data into the model was nevertheless proven for both examples.     

Multicriteria second‐order neural networks approach to imaging through turbulence

Atmospheric turbulence can greatly limit the spatial resolution in optical images obtained of space objects when imaged with ground‐based telescopes. Two widely used algorithms to remove atmospheric turbulence in this class of images are blind de‐convolution and speckle imaging. Both algorithms are effective in removing atmospheric turbulence, but they use different types of prior knowledge and have different strengths and weaknesses. We have developed a multicriteria cross entropy minimization approach to imaging through atmospheric turbulence and a second‐order neural network implementations. Our simulations illustrated the efficiency of our method. © 2003 Wiley Periodicals, Inc. Int J Imaging Syst Technol 13, 146–151, 2003; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ima.10037     

On the estimation of the expected number of accidents

We show that similar entities (drivers, intersections, bus companies, rail crossings) which in one period recorded "x" accidents do not have, on the average, "x" accidents in the subsequent period. The difference is large and systematic. This leads us to conclude that in circumstances in which only the safety estimates in these two periods matter, use of "x" to estimate the expected number of accidents has definite shortcomings. Better estimators are suggested, explored and their use is illustrated. We note that the suggested estimators are similar to what is used when estimation is based on a "treatment-control" type experimental design. It is hoped that the suggested estimators will alleviate some practical problems in the structuring of controlled experiments in safety research, eliminate bias-by-selection from uncontrolled studies and in general enhance the accuracy of safety estimates.     

人工神经网络的单调序列逼近

设Nn,Ф是以Ф为激活函数的具有n+1个神经元的前向单隐层人工神经网络的全体.主要证明了,若f∈C[0,1],则对于任意的ε>0,存在两个神经网络序列{Pn,Ф}和{Qn,Ф},使得在[0,1]上Qn,Ф(x)≤Qn+1,Ф(x)≤f(x)≤Pn+1,Ф(x)≤Pn,Ф(x),而且Pn,Ф(x)-Qn,Ф(x)≤(6+4(2~(1/2)))En,Ф(f),这里的En,Ф(f)为Nn,Ф中的元对f的最佳逼近.     

Super‐resolution of the undersampled and subpixel shifted image sequence by a neural network

Numerous approaches to super‐resolution (SR) of sequentially observed images (image sequence) of low resolution (LR) have been presented in the past two decades. However, neural network methods are almost ignored for solving SR problems. This is because the SR problem traditionally has been regarded as the optimization of an ill‐posed large set of linear equations. A designed neural network based on this has a large number of neurons, thereby requiring a long learning time. Also, the deduced cost function is overly complex. These defects limit applications of a neural network to an SR problem. We think that the underlying meaning of the SR problem should refer to super‐resolving an imaging system by image sequence observation, instead of merely improving the image sequence itself. SR can be regarded as a pattern mapping from LR to SR images. The parameters of the pattern mapping can be learned from the imaging process of the image sequence. This article presents a neural network for SR based on learning from the imaging process of the image sequence. In order to speed up the convergence, we employ vector mapping to train the neural network. A mapping vector is composed of some neighbor subpixels. Such a well‐trained neural network has powerful generalization ability so that it can be used directly to estimate the SR image of the other image sequences without learning again. Our simulations show the effectiveness of the proposed neural network. © 2004 Wiley Periodicals, Inc. Int J Imaging Syst Technol 14, 8–15, 2004; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ima.20001     

Nondestructive elastostatic identification of unilateral cracks through BEM and neural networks

An inverse problem in nonlinear elastostatics is considered which concerns the identification of unilateral contact cracks by means of boundary measurements for given static loadings. Highly nonlinear structural behaviour like closed cracks can hardly be identified. In this case, the analysis of more than one loading cases is proposed and tested in this paper. The direct problem is modelled by using a direct multiregion boundary element formulation. The arising linear complementarity problem is solved explicitly by a pivoting (Lemke) technique. In view of the complexity of the inverse problem, a neural network based identification approach is adopted which uses feed-forward multilayer neural networks trained by back-propagation, error-driven supervised training. The applicability of the method is demonstrated by some numerical examples.     

Maximum a posteriori estimation of spectral reflectance from color image and multipoint spectral measurements

Accurate color image reproduction under arbitrary illumination can be realized if the spectral reflectance functions in a scene are obtained. Although multispectral imaging is one of the promising methods to obtain the reflectance of a scene, it is expected to reduce the number of color channels without significant loss of accuracy. This paper presents what we believe to be a new method for estimating spectral reflectance functions from color image and multipoint spectral measurements based on maximum a posteriori (MAP) estimation. Multipoint spectral measurements are utilized as auxiliary information to improve the accuracy of spectral reflectance estimated from image data. Through simulations, it is confirmed that the proposed method improves the estimation accuracy, particularly when a scene includes subjects that belong to various categories.     

Optimizing the parameters of multilayered feedforward neural networks through Taguchi design of experiments

The size and training parameters of artificial neural networks have a critical effect on their performance. This paper presents the application of the Taguchi Design of Experiments (DoEs) off‐line quality control method in the optimization of the design parameters of a neural network. Being a ‘parallel’ approach, the method offers considerable benefits in time and accuracy when compared with the conventional serial approach of trial and error. The use of the Taguchi method ensures that the quality of the neural network is taken into account at the design stage. The interpretation of the experimental results is based on the statistical technique known as analysis of variance (ANOVA). The signal‐to‐noise ratio (S/N) is used in designing a robust neural network that is less sensitive to noise. The effect of design parameters and neural network behaviour are also revealed as a result. Although a Wood Veneer Inspection Neural Network (WVINN) is the particular application presented here, the design methodology can be applied to neural networks in general. Copyright © 2000 John Wiley & Sons, Ltd.     

Modelling the dependency of the Smith–Watson–Topper parameter on the cycles‐to‐failure using serial hybrid neural networks

Calculating the fatigue damage with a strain‐based approach requires an ?–N durability curve that links the strain amplitude to the corresponding number of cycles‐to‐failure. This ?–N curve is usually modelled by the Coffin–Manson relationship. If a loading mean‐level also needs to be considered, the original Coffin–Manson relationship is modified using a Smith–Watson–Topper parameter. In this article a methodology for modelling the dependence of the Smith–Watson–Topper parameter on the number of cycles‐to‐failure is presented. The core of the presented methodology represents a multilayer perceptron neural network combined with the Smith–Watson–Topper analytical model. The article presents the theoretical background of the methodology, which is applied for the case of the experimental fatigue data. The results show that it is possible to model ?–N curves for different influential parameters, such as the specimen's diameter and the testing temperature. The results further show that it is possible to predict ?–N curves even for those combinations of the influential parameters for which no experimental data about the material endurance is available. This fact makes the presented model very suitable for the application in an R&D process when a durability of a product should be estimated on the basis of a very limited set of experimental data about the material endurance characteristics.     

Pose estimation from a two-dimensional view by use of composite correlation filters and neural networks

We present a technique to estimate the pose of a three-dimensional object from a two-dimensional view. We first compute the correlation between the unknown image and several synthetic-discriminant-function filters constructed with known views of the object. We consider both linear and nonlinear correlations. The filters are constructed in such a way that the obtained correlation values depend on the pose parameters. We show that this dependence is not perfectly linear, in particular for nonlinear correlation. Therefore we use a two-layer neural network to retrieve the pose parameters from the correlation values. We demonstrate the technique by simultaneously estimating the in-plane and out-of-plane orientations of an airplane within an 8-deg portion. We show that a nonlinear correlation is necessary to identify the object and also to estimate its pose. On the other hand, linear correlation is more accurate and more robust. A combination of linear and nonlinear correlations gives the best results.     

On the pretreatment process for the object extraction in color image of wear debris

In this article, some pretreatment techniques used for the object extraction in color debris image were introduced, which was an important basic work in ferrographic technology to identify precisely wear debris produced by friction and wear from the relative motions between machine parts. These pretreatment techniques included image enhancement, image segmentation, filling pore, image erosion, and recognition of wear debris. The results showed that these methods were feasible and effective, which could be applied to extract object of color wear debris image successfully and precisely. It provided an important basis for the recognition technique of wear debris which was related to monitoring machine operation state and fault diagnosis. © 2008 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 17, 277–284, 2007     

Deep chest X‐ray: Detection and classification of lesions based on deep convolutional neural networks

We investigated whether a convolutional neural network (CNN) can enhance the usability of computer‐aided detection (CAD) of chest radiographs for various pulmonary abnormal lesions. The numbers of normal and abnormal patients were 6055 and 3463, respectively. Two radiologists delineated regions of interest for lesions and labeled the disease types as ground truths. The datasets were split into training, tuning, and testing as 7:1: 2. Total test sets were randomly selected in 1214 normal and 690 abnormal. A 5‐fold, cross‐validation was performed on our datasets. For the classification of normal and abnormal, we developed a CNN based on DenseNet169; for abnormal detection, The You Only Look Once (YOLO) v2 with DenseNet was used. Detection and classification of normal and five classes of diseases (nodule[s], consolidation, interstitial opacity, pleural effusion, and pneumothorax) on chest radiographs were analyzed. Our CNN model classified chest radiographs as normal or abnormal with an accuracy of 97.8%. For the results of the abnormal, F1 score, was 75.2 ± 2.28% for nodules, 55.0 ± 4.3% for consolidation, 78.2 ± 7.85% for interstitial opacity, 81.6 ± 2.07% for pleural effusion, and 70.0 ± 7.97% for pneumothorax, respectively. In addition, we conducted the experiments between our method and RetinaNet with only nodules. The results of our method and RetinaNet at cutoff‐0.5 in the free response operating characteristic curve were 83.45% and 80.55%, respectively. Our algorithm demonstrated viable detection and disease classification capacity and could be used for CAD of lung diseases on chest radiographs.     

An efficient and automatic glioblastoma brain tumor detection using shift‐invariant shearlet transform and neural networks

The detection and segmentation of tumor region in brain image is a critical task due to the similarity between abnormal and normal region. In this article, a computer‐aided automatic detection and segmentation of brain tumor is proposed. The proposed system consists of enhancement, transformation, feature extraction, and classification. The shift‐invariant shearlet transform (SIST) is used to enhance the brain image. Further, nonsubsampled contourlet transform (NSCT) is used as multiresolution transform which transforms the spatial domain enhanced image into multiresolution image. The texture features from grey level co‐occurrence matrix (GLCM), Gabor, and discrete wavelet transform (DWT) are extracted with the approximate subband of the NSCT transformed image. These extracted features are trained and classified into either normal or glioblastoma brain image using feed forward back propagation neural networks. Further, K‐means clustering algorithm is used to segment the tumor region in classified glioblastoma brain image. The proposed method achieves 89.7% of sensitivity, 99.9% of specificity, and 99.8% of accuracy.     

Near‐crack contour behaviour and extraction of log‐singular stress terms of the self‐regular traction boundary integral equation

This work contains an analytical study of the asymptotic near‐crack contour behaviour of stresses obtained from the self‐regular traction‐boundary integral equation (BIE), both in two and in three dimensions, and for various crack displacement modes. The flat crack case is chosen for detailed analysis of the singular stress for points approaching the crack contour. By imposing a condition of bounded stresses on the crack surface, the work shows that the boundary stresses on the crack are in fact zero for an unloaded crack, and the interior stresses reproduce the known inverse square root behaviour when the distance from the interior point to the crack contour approaches zero. The correct order of the stress singularity is obtained after the integrals for the self‐regular traction‐BIE formulation are evaluated analytically for the assumed displacement discontinuity model. Based on the analytic results, a new near‐crack contour self‐regular traction‐BIE is proposed for collocation points near the crack contour. In this new formulation, the asymptotic log‐singular stresses are identified and extracted from the BIE. Log‐singular stress terms are revealed for the free integrals written as contour integrals and for the self‐regularized integral with the integration region divided into sub‐regions. These terms are shown to cancel each other exactly when combined and can therefore be eliminated from the final BIE formulation. This work separates mathematical and physical singularities in a unique manner. Mathematical singularities are identified, and the singular information is all contained in the region near the crack contour. Copyright © 2003 John Wiley & Sons, Ltd.     

Effect of submicron admixtures on mechanical and self‐healing properties of cement‐based composites

This study presents the development of novel submicron super absorbent polymers (SAPs) used as admixtures in cement‐based matrices with significant advantages over conventional products. The produced SAPs were characterized in respect of their morphology and composition, while their water absorption capacity was determined in different electrolyte solutions. The hybrid core‐shell spherical structure of the fabricated materials offered significant compatibility enhancement with cement while the workability of the mixture was maintained. The assessment of the cement‐based composites including SAPs revealed that their flexural strength increased by 78%. Self‐healing/sealing behavior was assessed by monitoring the crack sealing via SEM, elemental analysis of the healing products, and determination of the water absorbance coefficient for different times of treatment. The cement/SAPs composites with a concentration of SAPs 2% by weight of cement exhibited self‐healing/sealing responsive capability when an artificial crack was induced. According to the SEM characterization, the crack demonstrated complete healing for the better part of its length after 28 days of treatment.     

Fatigue properties of self‐piercing riveted multi‐rivet joints in steel and aluminum sheets

Self‐piercing riveting (SPR) is an important joining technology for connecting steel and aluminum sheets. In this paper, AA6111 aluminum alloy and DP780 high‐strength steel were adopted to study the influence of fatigue on remaining static strength and energy absorption properties on self‐piercing riveting multi‐rivet joints. The results showed that energy absorption capacity of the specimens decreased significantly after high cycle fatigue. Fatigue reduced the remaining static lap shear strength of riveted specimens. Scanning electron microscopy (SEM) was used to analyze the cross section of fatigue specimens fractured by static tension. The results showed that fretting wear was found at the contact area between rivet and aluminum sheets. Fatigue bands and fatigue cracks appeared in fatigue specimens after high cycle fatigue, while those with low cycle fatigue specimens did not appear. Small cracks weaken the strength of the aluminum sheet, resulting in the static tensile strength of the riveted specimen with high cycle fatigue is lower than that of other fatigue specimens.     

Determination of S–N curves with the application of artificial neural networks

The present paper describes the application of artificial neural networks for estimating the finite-life fatigue strength and fatigue limit. A comprehensive database with results of single-stage tests on specimens which simulate structural components is evaluated and prepared for processing with the use of neural networks. The available data are subdivided into different classes. A total of six different data classes are specified. The results of the prediction by means of neural networks are superior to those obtained with conventional methods for calculating the fatigue strength. The experimental results are estimated with high accuracy.     

Fatigue life estimation of non‐penetrated butt weldments in ligth metals by artificial neural network approach

This study presents a model for estimating the fatigue life of magnesium and aluminium non‐penetrated butt‐welded joints using Artificial Neural Network (ANN). The input parameters for the network are stress concentration factor K_t and nominal stress amplitude s_a,n. The output parameter is the endurable number of load cycles N. Fatigue data were collected from the literature from three different sources. The experimental tests, on which the fatigue data are based, were carried out at the Fraunhofer Institute for Structural Durability and System Reliability (LBF), Darmstadt – Germany. The results determined with use of artificial neural network for welded magnesium and aluminium joints are displayed in the same scatter bands of SN‐lines. It is observed that the trained results are in good agreement with the tested data and artificial neural network is applicable for estimating the SN‐lines for non‐penetrated welded magnesium and aluminium joints under cyclic loading.