Search

Copyright of China Mechanical Engineering is the property of Editorial Board of China Mechanical Engineering and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In mobile ad-hoc networks (MANETs) always the quality-of-service (QoS) routing problem remains to exist and as a non-deterministic polynomial hard problem it is necessary to improve the QoS parameters to the most possible extent. For the considered MANET model, it is important to develop a suitable model that is capable of improving and enhancing the quality-of-service metrics. In this research study, a novel bio-inspired deep residual neural network (DResNet) architecture algorithm is developed over a MANET model for designing a most effective QoS routing protocol. The main aim of this proposed study is to locate a better routing path with satisfied QoS metrics for the MANET model and the bio-inspired deep learning neural model gets trained to meet the set fitness function. The new DResNet architecture operates with limited quantity of training data and its weights are optimized with novel IIWGSO technique. The proposed IIWGSO based DResNet model with the ad-hoc on-demand distance vector (AODV) protocol gets trained for the MANET model and its effectiveness is justified with QoS constraints being satisfied. The new IIWGSO-DResNet-AODV architecture resulted in better fitness value and QoS metrics proving its superiority compared to the existing techniques from previous literatures for the same MANET model. [ABSTRACT FROM AUTHOR]

This paper develops an approach for detecting landslide using IoT. The simulation of IoT is the preliminary step that helps to collect data. The suggested Water Particle Grey Wolf Optimization (WPGWO) is used for the routing. The Water Cycle Algorithm (WCA), Particle Swarm Optimization (PSO), and Grey Wolf Optimization (GWO) are combined in the suggested method (WPGWO). The fitness is newly modeled considering energy, link cost, distance, and delay. The maintenance of routes is done to assess the dependability of the network topology. The landslide detection process is carried out at the IoT base station. In feature selection, angular distance is used. Oversampling is used to enrich the data, and Deep Residual Network (DRN) — used for landslide identification — is trained using the proposed Water Cycle Particle Swarm Optimization (WCPSO) method, which combines WCA and PSO. The proposed WCPSO-based DRN offered effective performance with the highest energy of 0.049 J, throughput of 0.0495, accuracy of 95.7%, sensitivity of 97.2% and specificity of 93.9%. This approach demonstrated improved robustness and produced the global best optimal solution. For the proposed WPGWO, WCA, GWO, and PSO are linked to improve performance in determining the optimum routes. When comparing with existing methods the proposed WCPSO-based DRN offered effective performance. [ABSTRACT FROM AUTHOR]

The prediction of news popularity is having substantial importance for the digital advertisement community in terms of selecting and engaging users. Traditional approaches are based on empirical data collected through surveys and applied statistical measures to prove a hypothesis. However, predicting news popularity based on statistical measures applied to past data is highly questionable. Therefore, in this article, we predict news popularity using machine learning classification models and deep residual neural network models. Articles are usually made up of textual content and in many cases, images are also used. Although it is evident that the appropriate amount of textual data is required to extract features and create models, image data is also helpful in gaining useful information. In this article, we present a novel multimodal online news popularity prediction model based on ensemble learning. This research work acts as a guide for extensive feature engineering, feature extraction, feature selection and effective modelling to create a robust news popularity Prediction Model. Three kinds of features—meta‐features, text features and image features are used to design an influential and robust model. The relative error performance measure Root Mean Squared logarithmic error (RMSLE) is used to quantify the popularity prediction error. Further, the RMSLE outcome shows 0.351 which is the lowest error value given by the proposed model. Further, the most important features are also sought out to show the dependence of the best‐fit model on text and image features. [ABSTRACT FROM AUTHOR]

Dementia is an increasing global health challenge. Motoric Cognitive Risk Syndrome (MCR) is a predementia stage that can be used to predict future occurrence of dementia. Traditionally, gait speed and subjective memory complaints are used to identify older adults with MCR. Our previous studies indicated that dual-task upper-extremity motor performance (DTUEMP) quantified by a single wrist-worn sensor was correlated with both motor and cognitive function. Therefore, the DTUEMP had a potential to be used in the diagnosis of MCR. Instead of using inertial sensors to capture kinematic data of upper-extremity movements, here we proposed a deep neural network-based video processing model to obtain DTUEMP metrics from a 20-second repetitive elbow flexion-extension test under dual-task condition. In details, we used a deep residual neural network to obtain joint coordinate set of the elbow and wrist in each frame, and then used optical flow method to correct the joint coordinates generated by the neural network. The coordinate sets of all frames in a video recording were used to generate an angle sequence which represents rotation angle of the line between the wrist and elbow. Then, the DTUEMP metrics (the mean and SD of flexion and extension phase) were derived from angle sequences. Multi-task learning (MTL) was used to assess cognitive and motor function represented by MMSE and TUG scores based on DTUEMP metrics, with single-task learning (STL) linear model as a benchmark. The results showed a good agreement (r $\ge0.80$ and ICC $\ge0.58$ ) between the derived DTUEMP metrics from our proposed model and the ones from clinically validated sensor processing model. We also found that there were correlations with statistical significance (p < 0.05) between some of video-derived DTUEMP metrics (i.e. the mean of flexion time and extension time) and clinical cognitive scale (Mini-Mental State Examination, MMSE). Additionally, some of video-derived DTUEMP metrics (i.e. the mean and standard deviation of flexion time and extension time) were also associated with the scores of timed-up and go (TUG) which is a gold standard to measure functional mobility. Mean absolute percentage error (MAPE) of MTL surpassed that of STL (For MMSE, MTL: 18.63%, STL: 23.18%. For TUG, MTL: 17.88%, STL: 22.53%). The experiments with different light conditions and shot angles verified the robustness of our proposed video processing model to extract DTUEMP metrics in potentially various home environments (r $\ge0.58$ and ICC $\ge0.71$ ). This study shows possibility of replacing sensor processing model with video processing model for analyzing the DTUEMP and a promising future to adjuvant diagnosis of MCR via a mobile platform.

Constructing reliable and effective models to recognize human emotional states has become an important issue in recent years. In this article, we propose a double way deep residual neural network combined with brain network analysis, which enables the classification of multiple emotional states. To begin with, we transform the emotional EEG signals into five frequency bands by wavelet transform and construct brain networks by inter-channel correlation coefficients. These brain networks are then fed into a subsequent deep neural network block which contains several modules with residual connection and enhanced by channel attention mechanism and spatial attention mechanism. In the second way of the model, we feed the emotional EEG signals directly into another deep neural network block to extract temporal features. At the end of the two ways, the features are concatenated for classification. To verify the effectiveness of our proposed model, we carried out a series of experiments to collect emotional EEG from eight subjects. The average accuracy of the proposed model on our emotional dataset is 94.57%. In addition, the evaluation results on public databases SEED and SEED-IV are 94.55% and 78.91%, respectively, demonstrating the superiority of our model in emotion recognition tasks.

Globally, traffic accidents are of main concern because of more death rates and economic losses every year. Thus, road accident severity is the most important issue of concern, mainly in the undeveloped countries. Generally, traffic accidents result in severe human fatalities and large economic losses in real-world circumstances. Moreover, appropriate, precise prediction of traffic accidents has a high probability with regard to safeguarding public security as well as decreasing economic losses. Hence, the conventional accident prediction techniques are usually devised with statistical evaluations, which identify and evaluate the fundamental relationships among human variability, environmental aspects, traffic accidents and road geometry. However, the conventional approaches have major restrictions based on the assumptions regarding function kind and data distribution. In this paper, Aquila Anti-Coronavirus Optimization-based Deep Long Short-Term Memory (AACO-based Deep LSTM) is developed for road accident severity detection. Spearman's rank correlation coefficient and Deep Recurrent Neural Network (DRNN) are utilized for the feature fusion process. Data augmentation method is carried out to improve the detection performance. Deep LSTM detects the road accident and its severity, where Deep LSTM is trained by the designed AACO algorithm for better performance. The developed AACO-based Deep LSTM model outperformed other existing methods with the Mean Square Error (MSE), Root-Mean-Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) of 0.0145, 0.1204 and 0.075%, respectively. [ABSTRACT FROM AUTHOR]

Recently, Physics-informed neural networks (PINNs) have been widely applied to solving various types of partial differential equations (PDEs) such as Poisson equation, Klein–Gordon equation, and diffusion equation. However, it is difficult to obtain higher accurate solutions, especially at the boundary due to the gradient imbalance of different loss terms for the PINN model. In this work, an adaptive learning rate residual network algorithm based on physics-informed (adaptive-PIRN) is proposed to overcome this limitation of the PINN model. In the adaptive-PIRN model, an adaptive learning rate technique is introduced to adaptively configure appropriate weights to the residual loss of the governing equation and the loss of initial/boundary conditions (I/BCs) by utilizing gradient statistics, which can alleviate gradient imbalance of different loss terms in PINN. Besides, based on the idea of ResNet, the "short connection" technique is used in adaptive-PIRN model, which can ensure that the original information is identically mapped. This structure has stronger expressive capabilities than fully connected neural networks and can avoid gradient disappearance. Finally, three different types of PDE are conducted to demonstrate predictive accuracy of our model. In addition, it is clearly observed from the results that the adaptive-PIRN can balance the gradient of loss items to a great extent, which improves the effectiveness of this network. [ABSTRACT FROM AUTHOR]

Summary: An image processing and classification become an exciting research region for indicative probe because of the expansion of digital microscopic imaging. Acute lymphoblastic leukemia (ALL) is caused owing to broken in bone marrow, which is general among the people around the biosphere. A leukemia generally affects blood creating organs and hence, easily spread to other tissues by blood. The hematologists afflict more to differentiate leukemia existence in body of patients based on blood smears. Hence, in this article, Taylor political monarch butterfly optimization‐enabled deep residual neural network (Taylor PMBO‐enabled DRN) is devised for ALL classification. Segmentation is the process of separating the images into small pixels. A deep joint segmentation model effectively enhances the detection accuracy and it has better segmentation capability. For a better classification procedure, the important features, such as statistical, texture, and grid features, are extracted. For ALL detection and classification processes, a deep residual neural network (DRN) classifier is trained by the Taylor PMBO. The devised Taylor PMBO‐based DRN model obtained better performance than other method by testing accuracy, sensitivity, specificity, F1‐score, and precision of 0.9656, 0.9666, 0.9709, 0.9280, and 0.9321. [ABSTRACT FROM AUTHOR]

Constructing reliable and effective models to recognize human emotional states has become an important issue in recent years. In this article, we propose a double way deep residual neural network combined with brain network analysis, which enables the classification of multiple emotional states. To begin with, we transform the emotional EEG signals into five frequency bands by wavelet transform and construct brain networks by inter-channel correlation coefficients. These brain networks are then fed into a subsequent deep neural network block which contains several modules with residual connection and enhanced by channel attention mechanism and spatial attention mechanism. In the second way of the model, we feed the emotional EEG signals directly into another deep neural network block to extract temporal features. At the end of the two ways, the features are concatenated for classification. To verify the effectiveness of our proposed model, we carried out a series of experiments to collect emotional EEG from eight subjects. The average accuracy of the proposed model on our emotional dataset is 94.57%. In addition, the evaluation results on public databases SEED and SEED-IV are 94.55% and 78.91%, respectively, demonstrating the superiority of our model in emotion recognition tasks. [ABSTRACT FROM AUTHOR]

Dementia is an increasing global health challenge. Motoric Cognitive Risk Syndrome (MCR) is a predementia stage that can be used to predict future occurrence of dementia. Traditionally, gait speed and subjective memory complaints are used to identify older adults with MCR. Our previous studies indicated that dual-task upper-extremity motor performance (DTUEMP) quantified by a single wrist-worn sensor was correlated with both motor and cognitive function. Therefore, the DTUEMP had a potential to be used in the diagnosis of MCR. Instead of using inertial sensors to capture kinematic data of upper-extremity movements, here we proposed a deep neural network-based video processing model to obtain DTUEMP metrics from a 20-second repetitive elbow flexion-extension test under dual-task condition. In details, we used a deep residual neural network to obtain joint coordinate set of the elbow and wrist in each frame, and then used optical flow method to correct the joint coordinates generated by the neural network. The coordinate sets of all frames in a video recording were used to generate an angle sequence which represents rotation angle of the line between the wrist and elbow. Then, the DTUEMP metrics (the mean and SD of flexion and extension phase) were derived from angle sequences. Multi-task learning (MTL) was used to assess cognitive and motor function represented by MMSE and TUG scores based on DTUEMP metrics, with single-task learning (STL) linear model as a benchmark. The results showed a good agreement (r $\ge0.80$ and ICC $\ge0.58$) between the derived DTUEMP metrics from our proposed model and the ones from clinically validated sensor processing model. We also found that there were correlations with statistical significance (p < 0.05) between some of video-derived DTUEMP metrics (i.e. the mean of flexion time and extension time) and clinical cognitive scale (Mini-Mental State Examination, MMSE). Additionally, some of video-derived DTUEMP metrics (i.e. the mean and standard deviation of flexion time and extension time) were also associated with the scores of timed-up and go (TUG) which is a gold standard to measure functional mobility. Mean absolute percentage error (MAPE) of MTL surpassed that of STL (For MMSE, MTL: 18.63%, STL: 23.18%. For TUG, MTL: 17.88%, STL: 22.53%). The experiments with different light conditions and shot angles verified the robustness of our proposed video processing model to extract DTUEMP metrics in potentially various home environments (r $\ge0.58$ and ICC $\ge0.71$). This study shows possibility of replacing sensor processing model with video processing model for analyzing the DTUEMP and a promising future to adjuvant diagnosis of MCR via a mobile platform. [ABSTRACT FROM AUTHOR]

Current deep learning based image enhancement algorithms attempt to learn the mapping relationship between degraded images and clear images directly. These algorithms often ignore the fidelity constraint of the observational model. In order to improve the image enhancement performance, an improved deep residual neural network based image enhancement algorithm (DRNN-IE) for low dose CT images is proposed in this paper. DRNN-IE embeds the image enhancement task into a deep neural network, and achieves data consistency using multiple enhancement modules and back-projection modules. The enhancement modules in DRNN-IE produce new features through fusing low-level and high-level features. In order to improve the algorithm's generalization ability, a dual-parameter loss function is adopted to train and optimize the neural network. Experiments on real CT images show that the proposed algorithm has excellent enhancement performance and retains detailed information of low-dose CT images. [ABSTRACT FROM AUTHOR]

Synchronous motors play a significant role in a wide range of industrial applications. Their reliable operation is paramount. Any faults in synchronous motors can lead to costly downtime, decreased productivity, and potential safety hazards. By accurately diagnosing and classifying faults, we can proactively address issues before they escalate, ensuring the smooth operation of synchronous motors and minimizing the risk of equipment failure. The accurate diagnosis and fault detection in synchronous motors pose a significant challenge in their operation and maintenance. In the existing models, the feature data at various depths are not thoroughly extracted to maximize their feature extraction capability. Additionally, they employ a single support vector machine to make the final decision on the output. The single support vector machine may not consistently produce more accurate outcomes. Therefore, this paper proposes a novel fault diagnosis model based on a deep residual neural network and multiple support vector machines to diagnose mechanical and electrical faults of synchronous motors. The proposed model improves upon existing fault diagnosis models in two key aspects. Firstly, by employing a deep neural network, the model is able to effectively process and extract fault features from the motor fault dataset, capturing more nuanced information that may be missed by existing models. Secondly, the use of multiple support vector machines enhances the decision-making capability of the model, allowing for more accurate fault diagnosis. By combining these two aspects, the proposed model achieves superior diagnostic performance compared to single support vector machine-based models. Our proposed model has been rigorously evaluated using mechanical and electrical fault datasets, and the results of experimental tests clearly demonstrate its superior diagnostic performance when compared to existing fault diagnosis models. The synergy of deep neural networks and multiple support vector machines not only improves fault detection accuracy but also enhances the robustness and generalizability of the model, making it a valuable tool for real-world industrial applications. • A novel model is proposed for synchronous motor fault diagnostics. • A deep residual neural network effectively extracts fault features. • Multiple support vector machines enhance the detection capability. • Proposed model is compared with existing models. [ABSTRACT FROM AUTHOR]

Copyright of Hydrogeology Journal is the property of Springer Nature and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Point cloud segmentation is the premise and basis of many 3D perception tasks, such as intelligent driving, object detection and recognition, scene recognition and understanding. In this paper, we present an improved PointNet for 3D object part Segmentation, and named the proposed PointNet as Deep Residual Neural Network Based PointNet (DResNet-PointNet). The architecture of DResNet- PointNet was desigined based on the idea of residual networks. Residual networks can increase the depth of the DResNet-PointNet without network degradation. The depth of DResNet-PointNet is twice as deep as that of original PointNet model. Increasing the depth of DResNet-PointNet can improve its ability to express complex functions and generalization ability of complex classification problems, and achieve better approximation of complex functions, thus improving the accuracy of segmentation. The experimental results of part segmentation verify the feasibility and effectiveness of DResNet-PointNet. [ABSTRACT FROM AUTHOR]

Feature representation and discriminative learning are proven models and technologies in artificial intelligence fields; however, major challenges for machine learning on large biological datasets are learning an effective model with mechanistical explanation on the model determination and prediction. To satisfy such demands, we developed Vec2image, an explainable convolutional neural network framework for characterizing the feature engineering, feature selection and classifier training that is mainly based on the collaboration of principal component coordinate conversion, deep residual neural networks and embedded k-nearest neighbor representation on pseudo images of high-dimensional biological data, where the pseudo images represent feature measurements and feature associations simultaneously. Vec2image has achieved better performance compared with other popular methods and illustrated its efficiency on feature selection in cell marker identification from tissue-specific single-cell datasets. In particular, in a case study on type 2 diabetes (T2D) by multiple human islet scRNA-seq datasets, Vec2image first displayed robust performance on T2D classification model building across different datasets, then a specific Vec2image model was trained to accurately recognize the cell state and efficiently rank feature genes relevant to T2D which uncovered potential T2D cellular pathogenesis; and next the cell activity changes, cell composition imbalances and cell–cell communication dysfunctions were associated to our finding T2D feature genes from both population-shared and individual-specific perspectives. Collectively, Vec2image is a new and efficient explainable artificial intelligence methodology that can be widely applied in human-readable classification and prediction on the basis of pseudo image representation of biological deep sequencing data. [ABSTRACT FROM AUTHOR]

Rotating machinery is a very important mechanical device widely used in critical industrial applications. Efficient fault detection and diagnosis are key challenges in the maintenance and operational reliability of rotating machinery. To overcome this problem, a novel fault diagnosis method for rotating machinery based on deep residual neural network (DRNN) and data fusion is proposed. First, the time domain and frequency domain features of the original signal are extracted through the Short-time Fourier transform (STFT) layer, and then the deep residual network and the fusion embedding layer are used to fuse the time domain, frequency domain and spatial domain features to obtain high-quality low-dimensional fusion features. Finally, the fault type is obtained through the classifier. The proposed method is applied to the fault diagnosis of rolling bearing and gearbox, and the performance of the model has been tested comprehensively, including model training test, anti-noise test, fault tolerance test. The results confirm that the proposed method is much more effective and robust for feature learning, model training, anti-noise, fault tolerance and fault diagnosis than other fusion learning methods and single sensor-based methods. This fully reflects the advantages of multi-source information fusion in ensuring the reliable operation of rotating machinery. [ABSTRACT FROM AUTHOR]

Abstract Background Accurate prediction of protein structure is fundamentally important to understand biological function of proteins. Template-based modeling, including protein threading and homology modeling, is a popular method for protein tertiary structure prediction. However, accurate template-query alignment and template selection are still very challenging, especially for the proteins with only distant homologs available. Results We propose a new template-based modelling method called ThreaderAI to improve protein tertiary structure prediction. ThreaderAI formulates the task of aligning query sequence with template as the classical pixel classification problem in computer vision and naturally applies deep residual neural network in prediction. ThreaderAI first employs deep learning to predict residue-residue aligning probability matrix by integrating sequence profile, predicted sequential structural features, and predicted residue-residue contacts, and then builds template-query alignment by applying a dynamic programming algorithm on the probability matrix. We evaluated our methods both in generating accurate template-query alignment and protein threading. Experimental results show that ThreaderAI outperforms currently popular template-based modelling methods HHpred, CNFpred, and the latest contact-assisted method CEthreader, especially on the proteins that do not have close homologs with known structures. In particular, in terms of alignment accuracy measured with TM-score, ThreaderAI outperforms HHpred, CNFpred, and CEthreader by 56, 13, and 11%, respectively, on template-query pairs at the similarity of fold level from SCOPe data. And on CASP13’s TBM-hard data, ThreaderAI outperforms HHpred, CNFpred, and CEthreader by 16, 9 and 8% in terms of TM-score, respectively. Conclusions These results demonstrate that with the help of deep learning, ThreaderAI can significantly improve the accuracy of template-based structure prediction, especially for distant-homology proteins.

Background: Accurate prediction of protein structure is fundamentally important to understand biological function of proteins. Template-based modeling, including protein threading and homology modeling, is a popular method for protein tertiary structure prediction. However, accurate template-query alignment and template selection are still very challenging, especially for the proteins with only distant homologs available. Results: We propose a new template-based modelling method called ThreaderAI to improve protein tertiary structure prediction. ThreaderAI formulates the task of aligning query sequence with template as the classical pixel classification problem in computer vision and naturally applies deep residual neural network in prediction. ThreaderAI first employs deep learning to predict residue-residue aligning probability matrix by integrating sequence profile, predicted sequential structural features, and predicted residue-residue contacts, and then builds template-query alignment by applying a dynamic programming algorithm on the probability matrix. We evaluated our methods both in generating accurate template-query alignment and protein threading. Experimental results show that ThreaderAI outperforms currently popular template-based modelling methods HHpred, CNFpred, and the latest contact-assisted method CEthreader, especially on the proteins that do not have close homologs with known structures. In particular, in terms of alignment accuracy measured with TM-score, ThreaderAI outperforms HHpred, CNFpred, and CEthreader by 56, 13, and 11%, respectively, on template-query pairs at the similarity of fold level from SCOPe data. And on CASP13's TBM-hard data, ThreaderAI outperforms HHpred, CNFpred, and CEthreader by 16, 9 and 8% in terms of TM-score, respectively. Conclusions: These results demonstrate that with the help of deep learning, ThreaderAI can significantly improve the accuracy of template-based structure prediction, especially for distant-homology proteins. [ABSTRACT FROM AUTHOR]

Aiming at the image features of stock data, considering the picture features of stock data and the characteristics of CNN's good at extracting picture features, the paper proposed a stock price trend prediction model CNN-M based on a Convolutional Neural Network. At the same time, based on the excellent image feature extraction ability of the residual network, this paper proposed a residual network-based stock price trend prediction model ResNet-M based on the Conventional Neural Network. The experimental results showed that the prediction ability of the improved residual network-based prediction model Resnet-M is superior to the CNN model. [ABSTRACT FROM AUTHOR]

Purpose: To develop a deep learning approach based on deep residual neural network (ResNet101) for the automated detection of glaucomatous optic neuropathy (GON) using color fundus images, understand the process by which the model makes predictions, and explore the effect of the integration of fundus images and the medical history data from patients. Methods: A total of 34,279 fundus images and the corresponding medical history data were retrospectively collected from cohorts of 2371 adult patients, and these images were labeled by 8 glaucoma experts, in which 26,585 fundus images (12,618 images with GON-confirmed eyes, 1114 images with GON-suspected eyes, and 12,853 NORMAL eye images) were included. We adopted 10-fold cross-validation strategy to train and optimize our model. This model was tested in an independent testing dataset consisting of 3481 images (1524 images from NORMAL eyes, 1442 images from GON-confirmed eyes, and 515 images from GON-suspected eyes) from 249 patients. Moreover, the performance of the best model was compared with results obtained by two experts. Accuracy, sensitivity, specificity, kappa value, and area under receiver operating characteristic (AUC) were calculated. Further, we performed qualitative evaluation of model predictions and occlusion testing. Finally, we assessed the effect of integrating medical history data in the final classification. Results: In a multiclass comparison between GON-confirmed eyes, GON-suspected eyes and NORMAL eyes, our model achieved 0.941 (95% confidence interval [CI], 0.936–0.946) accuracy, 0.957 (95% CI, 0.953–0.961) sensitivity, and 0.929 (95% CI, 0.923–0.935) specificity. The AUC distinguishing referrals (GON-confirmed and GON-suspected eyes) from observation was 0.992 (95% CI, 0.991–0.993). Our best model had a kappa value of 0.927, while the two experts' kappa values were 0.928 and 0.925 independently. The best 2 binary classifiers distinguishing GON-confirmed/GON-suspected eyes from NORMAL eyes obtained 0.955, 0.965 accuracy, 0.977, 0.998 sensitivity, and 0.929, 0.954 specificity, while the AUC was 0.992, 0.999 respectively. Additionally, the occlusion testing showed that our model identified the neuroretinal rim region, retinal nerve fiber layer (RNFL) defect areas (superior or inferior) as the most important parts for the discrimination of GON, which evaluated fundus images in a way similar to clinicians. Finally, the results of integration of fundus images with medical history data showed a slight improvement in sensitivity and specificity with similar AUCs. Conclusions: This approach could discriminate GON with high accuracy, sensitivity, specificity, and AUC using color fundus photographs. It may provide a second opinion on the diagnosis of glaucoma to the specialist quickly, efficiently and at low cost, and assist doctors and the public in large-scale screening for glaucoma. [ABSTRACT FROM AUTHOR]

Use of simulation data is necessary for training fault diagnostic models because there is an insufficient amount of fault data available for intricate supercritical flexible shafting. A hybrid dynamic modelling approach combining finite element and lumped mass techniques was used to construct dynamic models of the system in both normal and fault states. The simulation signals corresponding to each state were obtained through numerical calculations and subsequently compared with the existing literature to ensure the accuracy and validity of the dynamic model. By establishing this foundation, dependable training data can be acquired for fault diagnosis within a system. A deep residual neural network with a multi-scale convolution kernel (MSResNet) was used to conduct fault diagnosis of the flexible shafting. The efficacy of the suggested approach was substantiated through an experimental analysis. The outcomes of this research establish a theoretical foundation for fault diagnosis of flexible shafting in scenarios with an insufficient number of fault samples. • The dynamic models of the normal and fault states of the flexible shafting were established. • A novel diagnosis method to detect faults in the flexible shafting using dynamics simulation and deep learning. • The proposed method was validated through experiments on the flexible shafting test rig. [ABSTRACT FROM AUTHOR]

Short-term load forecasting is integral to the energy planning sector. Various techniques have been employed to achieve effective operation of power systems and efficient market management. We present a scalable system for day-ahead household electrical energy consumption forecasting, named HousEEC. The proposed forecasting method is based on a deep residual neural network, and integrates multiple sources of information by extracting features from (i) contextual data (weather, calendar), and (ii) the historical load of the particular household and all households present in the dataset. Additionally, we compute novel domain-specific time-series features that allow the system to better model the pattern of energy consumption of the household. The experimental analysis and evaluation were performed on one of the most extensive datasets for household electrical energy consumption, Pecan Street, containing almost four years of data. Multiple test cases show that the proposed model provides accurate load forecasting results, achieving a root-mean-square error score of 0.44 kWh and mean absolute error score of 0.23 kWh, for short-term load forecasting for 300 households. The analysis showed that, for hourly forecasting, our model had 8% error (22 kWh), which is 4 percentage points better than the benchmark model. The daily analysis showed that our model had 2% error (131 kWh), which is significantly less compared to the benchmark model, with 6% error (360 kWh).

Developing accurate, efficient, and robust closure models is essential in the construction of reduced order models (ROMs) for realistic nonlinear systems, which generally require drastic ROM mode truncations. We propose a deep residual neural network (ResNet) closure learning framework for ROMs of nonlinear systems. The novel ResNet-ROM framework consists of two steps: (i) In the first step, we use ROM projection to filter the given nonlinear system and construct a spatially filtered ROM. This filtered ROM is low-dimensional, but is not closed. (ii) In the second step, we use ResNet to close the filtered ROM, i.e., to model the interaction between the resolved and unresolved ROM modes. We emphasize that in the new ResNet-ROM framework, data is used only to complement classical physical modeling (i.e., only in the closure modeling component), not to completely replace it. We also note that the new ResNet-ROM is built on general ideas of spatial filtering and deep learning and is independent of (restrictive) phenomenological arguments, e.g., of eddy viscosity type. The numerical experiments for the 1D Burgers equation show that the ResNet-ROM is significantly more accurate than the standard projection ROM. The new ResNet-ROM is also more accurate and significantly more efficient than other modern ROM closure models.

• A deep learning model based on multi-frequency domain information is proposed for rotating machinery diagnosis. • The diagnostic framework has fast processing speed and does not require tedious parameter adjustment process. • Generalization and anti-noise ability are improved by combining ResNet and multi-sensor data fusion. • Abundant fault simulation experiments capture common rotating machinery failures to provide data support. Rotating machinery is a key component of nuclear power plants (NPPs). The integrity of rotating machine is related to the safety and economy of the entire NPPs. In order to achieve better and more robust diagnostic performance, this work proposes an intelligent fault diagnosis method based on multi-sensor and deep residual neural network. This method gives full play to the value of multi-sensor information, and utilizes the powerful learning ability of deep learning model to realize the identification of rotating machinery fault types. The effectiveness of the method is evaluated by using the motor dataset and the bearing dataset from fault simulation experiment bench. In addition, the anti-noise ability of the method is tested and compared with other methods. The results show that the proposed method has higher diagnosis accuracy and stronger robustness, demonstrating the potential application value in rotating machinery of NPPs. [ABSTRACT FROM AUTHOR]

Colorectal cancers may occur in colon region of human body because of late detection of polyps. Therefore, colonoscopists often use colonoscopy device to view the entire colon in their routine practice to remove polyps by excisional biopsy. The aim of this study is to develop a new imbalance-aware loss function, i.e., omni-comprehensive loss, to be used in deep neural networks to overcome both imbalanced dataset and the vanishing gradient problem in identifying the related regions of a polyp. Another reason of developing a new loss function is to be able to produce a more comprehensive one that has evaluation capabilities of region-based, shape-aware, and pixel-wise distribution loss approaches at once. To measure the performance of the new loss function, two scenarios have been conducted. First, an 18-layer residual network as backbone with UNet as the decoder is implemented. Second, a 34-layer residual network as the encoder and a UNet as the decoder is designed. For both scenarios, the results of using popular imbalance-aware losses are compared with those of using our proposed new loss function. During training and 5-fold cross validation steps, multiple publicly available datasets are used. In addition to original data in these datasets, their augmented versions are also created by flipping, scaling, rotating and contrast-limited adaptive histogram equalization operations. As a result, our proposed new custom loss function produced the best performance metrics compared with the popular loss functions., Competing Interests: Declaration of competing interest The authors declare that this article content has no conflict of interest., (Copyright © 2022 Elsevier Ltd. All rights reserved.)

• We developed a novel deep learning network to segment the gastrocnemius and soleus in shank ultrasound (US) images. • As the sublayer unit of each layer of a U-net, deep Resnet was used to meet the challenges of poor US image quality and low contrast. • Dilated convolution was used to extract deeper feature information, and then reduce the interference of a large number of shadows. • This method has achieved a mean DSC of the Gas and Sol muscles of 94.82% and 90.72%, which outperforms other state-of-the-art methods. Segmentation of the shank gastrocnemius (Gas) and soleus (Sol) muscles in ultrasound (US) images allows to extract the muscle features, which are important for the early diagnosis of muscle atrophy. The automatic segmentation of the muscles is a challenging task, and deep learning (DL) provides a solution to this problem, which can effectively extract representative features from the muscle regions and background of the images. In this study, we propose ResTU-net, an automatic segmentation method based on improved U-net network, to segment the Gas and Sol muscles in shank US images. This network uses the deep residual neural network (Resnet) as the sublayer unit of each layer of a U-net, and can effectively combine the features of each layer with those of the next layer to meet the challenges of poor US image quality and low contrast. In addition, dilated convolution is used instead of the pooling layer in the network to prevent information loss during training. Experiments were performed on 3350 shank US images from 23 Sprague Dawley (SD) rats, among them, 2650 shank US images were used for network training and 700 for network validation. Compared with state-of-the-art networks, the experimental results show that the method can achieved the best segmentation capability results and a mean Dice similarity coefficient (DSC) of the Gas and Sol muscles of 94.82% and 90.72%, respectively. This work indicates that the proposed fully automatic segmentation method may be accurately and efficiently applied to Gas and Sol muscles segmentation in shank US images. [ABSTRACT FROM AUTHOR]

Uneven dust accumulation can significantly influence the thermal balance between different regions of photovoltaic (PV) panels, leading to a sharp decrease in power generation efficiency and service life. In this paper, a new identification method for uneven dust accumulation on the surface of PV panels is developed to analyze the dust state (concentration and distribution) quantitatively. First, a novel deep residual neural network (DRNN) is proposed to obtain the regional dust concentration. The residual elements in the model are connected in the form of skipping layers to reduce the order of the weight matrices and improve the network flexibility and the feature learning accuracy. Second, an image preprocessing method is designed to classify the dust accumulation. It includes transformation and correction, removal of the silver grid, nonlinear interpolation, equivalent segmentation, and clustering. A new error evaluation method, error loop, is proposed to analyze the consistency between the measured and the experimental results. The results show that the DRNN has better prediction accuracy than other methods. The R2 and mean absolute error (MAE) of the DRNN are 78.7% and 3.67, respectively. In addition, three conditions are used to verify the performance of the identification method for determining uneven dust accumulation. The average evaluation coefficients of the error loop are 1.19, 0.77, and 1.10, respectively, meeting the design requirements. The proposed method can provide theoretical support for the intelligent operation and maintenance of PV systems. • An identification method for the uneven dust of PV panels is proposed. • A DRNN network architecture of free identity mapping is established. • A multiscale error evaluation method (error loop) is presented. • The model is verified by experiments and error loop evaluation. [ABSTRACT FROM AUTHOR]

Short-term load forecasting is integral to the energy planning sector. Various techniques have been employed to achieve effective operation of power systems and efficient market management. We present a scalable system for day-ahead household electrical energy consumption forecasting, named HousEEC. The proposed forecasting method is based on a deep residual neural network, and integrates multiple sources of information by extracting features from (i) contextual data (weather, calendar), and (ii) the historical load of the particular household and all households present in the dataset. Additionally, we compute novel domain-specific time-series features that allow the system to better model the pattern of energy consumption of the household. The experimental analysis and evaluation were performed on one of the most extensive datasets for household electrical energy consumption, Pecan Street, containing almost four years of data. Multiple test cases show that the proposed model provides accurate load forecasting results, achieving a root-mean-square error score of 0.44 kWh and mean absolute error score of 0.23 kWh, for short-term load forecasting for 300 households. The analysis showed that, for hourly forecasting, our model had 8% error (22 kWh), which is 4 percentage points better than the benchmark model. The daily analysis showed that our model had 2% error (131 kWh), which is significantly less compared to the benchmark model, with 6% error (360 kWh). [ABSTRACT FROM AUTHOR]