Handling missing data accurately is critical in clinical research, where data quality directly impacts decision-making and patient outcomes. While deep learning (DL) techniques for data imputation have gained attentio...Handling missing data accurately is critical in clinical research, where data quality directly impacts decision-making and patient outcomes. While deep learning (DL) techniques for data imputation have gained attention, challenges remain, especially when dealing with diverse data types. In this study, we introduce a novel data imputation method based on a modified convolutional neural network, specifically, a Deep Residual-Convolutional Neural Network (DRes-CNN) architecture designed to handle missing values across various datasets. Our approach demonstrates substantial improvements over existing imputation techniques by leveraging residual connections and optimized convolutional layers to capture complex data patterns. We evaluated the model on publicly available datasets, including Medical Information Mart for Intensive Care (MIMIC-III and MIMIC-IV), which contain critical care patient data, and the Beijing Multi-Site Air Quality dataset, which measures environmental air quality. The proposed DRes-CNN method achieved a root mean square error (RMSE) of 0.00006, highlighting its high accuracy and robustness. We also compared with Low Light-Convolutional Neural Network (LL-CNN) and U-Net methods, which had RMSE values of 0.00075 and 0.00073, respectively. This represented an improvement of approximately 92% over LL-CNN and 91% over U-Net. The results showed that this DRes-CNN-based imputation method outperforms current state-of-the-art models. These results established DRes-CNN as a reliable solution for addressing missing data.展开更多
In the burgeoning field of anomaly detection within attributed networks,traditional methodologies often encounter the intricacies of network complexity,particularly in capturing nonlinearity and sparsity.This study in...In the burgeoning field of anomaly detection within attributed networks,traditional methodologies often encounter the intricacies of network complexity,particularly in capturing nonlinearity and sparsity.This study introduces an innovative approach that synergizes the strengths of graph convolutional networks with advanced deep residual learning and a unique residual-based attention mechanism,thereby creating a more nuanced and efficient method for anomaly detection in complex networks.The heart of our model lies in the integration of graph convolutional networks that capture complex structural relationships within the network data.This is further bolstered by deep residual learning,which is employed to model intricate nonlinear connections directly from input data.A pivotal innovation in our approach is the incorporation of a residual-based attention mech-anism.This mechanism dynamically adjusts the importance of nodes based on their residual information,thereby significantly enhancing the sensitivity of the model to subtle anomalies.Furthermore,we introduce a novel hypersphere mapping technique in the latent space to distinctly separate normal and anomalous data.This mapping is the key to our model’s ability to pinpoint anomalies with greater precision.An extensive experimental setup was used to validate the efficacy of the proposed model.Using attributed social network datasets,we demonstrate that our model not only competes with but also surpasses existing state-of-the-art methods in anomaly detection.The results show the exceptional capability of our model to handle the multifaceted nature of real-world networks.展开更多
A deep-sea riser is a crucial component of the mining system used to lift seafloor mineral resources to the vessel.Even minor damage to the riser can lead to substantial financial losses,environmental impacts,and safe...A deep-sea riser is a crucial component of the mining system used to lift seafloor mineral resources to the vessel.Even minor damage to the riser can lead to substantial financial losses,environmental impacts,and safety hazards.However,identifying modal parameters for structural health monitoring remains a major challenge due to its large deformations and flexibility.Vibration signal-based methods are essential for detecting damage and enabling timely maintenance to minimize losses.However,accurately extracting features from one-dimensional(1D)signals is often hindered by various environmental factors and measurement noises.To address this challenge,a novel approach based on a residual convolutional auto-encoder(RCAE)is proposed for detecting damage in deep-sea mining risers,incorporating a data fusion strategy.First,principal component analysis(PCA)is applied to reduce environmental fluctuations and fuse multisensor strain readings.Subsequently,a 1D-RCAE is used to extract damage-sensitive features(DSFs)from the fused dataset.A Mahalanobis distance indicator is established to compare the DSFs of the testing and healthy risers.The specific threshold for these distances is determined using the 3σcriterion,which is employed to assess whether damage has occurred in the testing riser.The effectiveness and robustness of the proposed approach are verified through numerical simulations of a 500-m riser and experimental tests on a 6-m riser.Moreover,the impact of contaminated noise and environmental fluctuations is examined.Results show that the proposed PCA-1D-RCAE approach can effectively detect damage and is resilient to measurement noise and environmental fluctuations.The accuracy exceeds 98%under noise-free conditions and remains above 90%even with 10 dB noise.This novel approach has the potential to establish a new standard for evaluating the health and integrity of risers during mining operations,thereby reducing the high costs and risks associated with failures.Maintenance activities can be scheduled more efficiently by enabling early and accurate detection of riser damage,minimizing downtime and avoiding catastrophic failures.展开更多
In daily life,keyword spotting plays an important role in human-computer interaction.However,noise often interferes with the extraction of time-frequency information,and achieving both computational efficiency and rec...In daily life,keyword spotting plays an important role in human-computer interaction.However,noise often interferes with the extraction of time-frequency information,and achieving both computational efficiency and recognition accuracy on resource-constrained devices such as mobile terminals remains a major challenge.To address this,we propose a novel time-frequency dual-branch parallel residual network,which integrates a Dual-Branch Broadcast Residual module and a Time-Frequency Coordinate Attention module.The time-domain and frequency-domain branches are designed in parallel to independently extract temporal and spectral features,effectively avoiding the potential information loss caused by serial stacking,while enhancing information flow and multi-scale feature fusion.In terms of training strategy,a curriculum learning approach is introduced to progressively improve model robustness fromeasy to difficult tasks.Experimental results demonstrate that the proposed method consistently outperforms existing lightweight models under various signal-to-noise ratio(SNR)conditions,achieving superior far-field recognition performance on the Google Speech Commands V2 dataset.Notably,the model maintains stable performance even in low-SNR environments such as–10 dB,and generalizes well to unseen SNR conditions during training,validating its robustness to novel noise scenarios.Furthermore,the proposed model exhibits significantly fewer parameters,making it highly suitable for deployment on resource-limited devices.Overall,the model achieves a favorable balance between performance and parameter efficiency,demonstrating strong potential for practical applications.展开更多
Abnormal network traffic, as a frequent security risk, requires a series of techniques to categorize and detect it. Existing network traffic anomaly detection still faces challenges: the inability to fully extract loc...Abnormal network traffic, as a frequent security risk, requires a series of techniques to categorize and detect it. Existing network traffic anomaly detection still faces challenges: the inability to fully extract local and global features, as well as the lack of effective mechanisms to capture complex interactions between features;Additionally, when increasing the receptive field to obtain deeper feature representations, the reliance on increasing network depth leads to a significant increase in computational resource consumption, affecting the efficiency and performance of detection. Based on these issues, firstly, this paper proposes a network traffic anomaly detection model based on parallel dilated convolution and residual learning (Res-PDC). To better explore the interactive relationships between features, the traffic samples are converted into two-dimensional matrix. A module combining parallel dilated convolutions and residual learning (res-pdc) was designed to extract local and global features of traffic at different scales. By utilizing res-pdc modules with different dilation rates, we can effectively capture spatial features at different scales and explore feature dependencies spanning wider regions without increasing computational resources. Secondly, to focus and integrate the information in different feature subspaces, further enhance and extract the interactions among the features, multi-head attention is added to Res-PDC, resulting in the final model: multi-head attention enhanced parallel dilated convolution and residual learning (MHA-Res-PDC) for network traffic anomaly detection. Finally, comparisons with other machine learning and deep learning algorithms are conducted on the NSL-KDD and CIC-IDS-2018 datasets. The experimental results demonstrate that the proposed method in this paper can effectively improve the detection performance.展开更多
Microphone array-based sound source localization(SSL)is widely used in a variety of occasions such as video conferencing,robotic hearing,speech enhancement,speech recognition and so on.The traditional SSL methods cann...Microphone array-based sound source localization(SSL)is widely used in a variety of occasions such as video conferencing,robotic hearing,speech enhancement,speech recognition and so on.The traditional SSL methods cannot achieve satisfactory performance in adverse noisy and reverberant environments.In order to improve localization performance,a novel SSL algorithm using convolutional residual network(CRN)is proposed in this paper.The spatial features including time difference of arrivals(TDOAs)between microphone pairs and steered response power-phase transform(SRPPHAT)spatial spectrum are extracted in each Gammatone sub-band.The spatial features of different sub-bands with a frame are combine into a feature matrix as the input of CRN.The proposed algorithm employ CRN to fuse the spatial features.Since the CRN introduces the residual structure on the basis of the convolutional network,it reduce the difficulty of training procedure and accelerate the convergence of the model.A CRN model is learned from the training data in various reverberation and noise environments to establish the mapping regularity between the input feature and the sound azimuth.Through simulation verification,compared with the methods using traditional deep neural network,the proposed algorithm can achieve a better localization performance in SSL task,and provide better generalization capacity to untrained noise and reverberation.展开更多
As a common and high-risk type of disease,heart disease seriously threatens people’s health.At the same time,in the era of the Internet of Thing(IoT),smart medical device has strong practical significance for medical...As a common and high-risk type of disease,heart disease seriously threatens people’s health.At the same time,in the era of the Internet of Thing(IoT),smart medical device has strong practical significance for medical workers and patients because of its ability to assist in the diagnosis of diseases.Therefore,the research of real-time diagnosis and classification algorithms for arrhythmia can help to improve the diagnostic efficiency of diseases.In this paper,we design an automatic arrhythmia classification algorithm model based on Convolutional Neural Network(CNN)and Encoder-Decoder model.The model uses Long Short-Term Memory(LSTM)to consider the influence of time series features on classification results.Simultaneously,it is trained and tested by the MIT-BIH arrhythmia database.Besides,Generative Adversarial Networks(GAN)is adopted as a method of data equalization for solving data imbalance problem.The simulation results show that for the inter-patient arrhythmia classification,the hybrid model combining CNN and Encoder-Decoder model has the best classification accuracy,of which the accuracy can reach 94.05%.Especially,it has a better advantage for the classification effect of supraventricular ectopic beats(class S)and fusion beats(class F).展开更多
Noise reduction analysis of signals is essential for modern underwater acoustic detection systems.The traditional noise reduction techniques gradually lose efficacy because the target signal is masked by biological an...Noise reduction analysis of signals is essential for modern underwater acoustic detection systems.The traditional noise reduction techniques gradually lose efficacy because the target signal is masked by biological and natural noise in the marine environ-ment.The feature extraction method combining time-frequency spectrograms and deep learning can effectively achieve the separation of noise and target signals.A fully convolutional encoder-decoder neural network(FCEDN)is proposed to address the issue of noise reduc-tion in underwater acoustic signals.The time-domain waveform map of underwater acoustic signals is converted into a wavelet low-frequency analysis recording spectrogram during the denoising process to preserve as many underwater acoustic signal characteristics as possible.The FCEDN is built to learn the spectrogram mapping between noise and target signals that can be learned at each time level.The transposed convolution transforms are introduced,which can transform the spectrogram features of the signals into listenable audio files.After evaluating the systems on the ShipsEar Dataset,the proposed method can increase SNR and SI-SNR by 10.02 and 9.5dB,re-spectively.展开更多
Amodel that can obtain rapid and accurate detection of coronavirus disease 2019(COVID-19)plays a significant role in treating and preventing the spread of disease transmission.However,designing such amodel that can ba...Amodel that can obtain rapid and accurate detection of coronavirus disease 2019(COVID-19)plays a significant role in treating and preventing the spread of disease transmission.However,designing such amodel that can balance the detection accuracy andweight parameters ofmemorywell to deploy a mobile device is challenging.Taking this point into account,this paper fuses the convolutional neural network and residual learning operations to build a multi-class classification model,which improves COVID-19 pneumonia detection performance and keeps a trade-off between the weight parameters and accuracy.The convolutional neural network can extract the COVID-19 feature information by repeated convolutional operations.The residual learning operations alleviate the gradient problems caused by stacking convolutional layers and enhance the ability of feature extraction.The ability further enables the proposed model to acquire effective feature information at a lowcost,which canmake ourmodel keep smallweight parameters.Extensive validation and comparison with other models of COVID-19 pneumonia detection on the well-known COVIDx dataset show that(1)the sensitivity of COVID-19 pneumonia detection is improved from 88.2%(non-COVID-19)and 77.5%(COVID-19)to 95.3%(non-COVID-19)and 96.5%(COVID-19),respectively.The positive predictive value is also respectively increased from72.8%(non-COVID-19)and 89.0%(COVID-19)to 88.8%(non-COVID-19)and 95.1%(COVID-19).(2)Compared with the weight parameters of the COVIDNet-small network,the value of the proposed model is 13 M,which is slightly higher than that(11.37 M)of the COVIDNet-small network.But,the corresponding accuracy is improved from 85.2%to 93.0%.The above results illustrate the proposed model can gain an efficient balance between accuracy and weight parameters.展开更多
Robotic grasps play an important role in the service and industrial fields,and the robotic arm can grasp the object properly depends on the accuracy of the grasping detection result.In order to predict grasping detect...Robotic grasps play an important role in the service and industrial fields,and the robotic arm can grasp the object properly depends on the accuracy of the grasping detection result.In order to predict grasping detection positions for known or unknown objects by a modular robotic system,a convolutional neural network(CNN)with the residual block is proposed,which can be used to generate accurate grasping detection for input images of the scene.The proposed model architecture was trained on the standard Cornell grasp dataset and evaluated on the test dataset.Moreover,it was evaluated on different types of household objects and cluttered multi-objects.On the Cornell grasp dataset,the accuracy of the model on image-wise splitting detection and object-wise splitting detection achieved 95.5%and 93.6%,respectively.Further,the real detection time per image was 109 ms.The experimental results show that the model can quickly detect the grasping positions of a single object or multiple objects in image pixels in real time,and it keeps good stability and robustness.展开更多
Deep convolutional neural networks(CNNs)have demonstrated remarkable performance in video super-resolution(VSR).However,the ability of most existing methods to recover fine details in complex scenes is often hindered ...Deep convolutional neural networks(CNNs)have demonstrated remarkable performance in video super-resolution(VSR).However,the ability of most existing methods to recover fine details in complex scenes is often hindered by the loss of shallow texture information during feature extraction.To address this limitation,we propose a 3D Convolutional Enhanced Residual Video Super-Resolution Network(3D-ERVSNet).This network employs a forward and backward bidirectional propagation module(FBBPM)that aligns features across frames using explicit optical flow through lightweight SPyNet.By incorporating an enhanced residual structure(ERS)with skip connections,shallow and deep features are effectively integrated,enhancing texture restoration capabilities.Furthermore,3D convolution module(3DCM)is applied after the backward propagation module to implicitly capture spatio-temporal dependencies.The architecture synergizes these components where FBBPM extracts aligned features,ERS fuses hierarchical representations,and 3DCM refines temporal coherence.Finally,a deep feature aggregation module(DFAM)fuses the processed features,and a pixel-upsampling module(PUM)reconstructs the high-resolution(HR)video frames.Comprehensive evaluations on REDS,Vid4,UDM10,and Vim4 benchmarks demonstrate well performance including 30.95 dB PSNR/0.8822 SSIM on REDS and 32.78 dB/0.8987 on Vim4.3D-ERVSNet achieves significant gains over baselines while maintaining high efficiency with only 6.3M parameters and 77ms/frame runtime(i.e.,20×faster than RBPN).The network’s effectiveness stems from its task-specific asymmetric design that balances explicit alignment and implicit fusion.展开更多
The fifth-generation (5G) communication requires a highly accurate estimation of the channel state information (CSI)to take advantage of the massive multiple-input multiple-output(MIMO) system. However, traditional ch...The fifth-generation (5G) communication requires a highly accurate estimation of the channel state information (CSI)to take advantage of the massive multiple-input multiple-output(MIMO) system. However, traditional channel estimation methods do not always yield reliable estimates. The methodology of this paper consists of deep residual shrinkage network (DRSN)neural network-based method that is used to solve this problem.Thus, the channel estimation approach, based on DRSN with its learning ability of noise-containing data, is first introduced. Then,the DRSN is used to train the noise reduction process based on the results of the least square (LS) channel estimation while applying the pilot frequency subcarriers, where the initially estimated subcarrier channel matrix is considered as a three-dimensional tensor of the DRSN input. Afterward, a mixed signal to noise ratio (SNR) training data strategy is proposed based on the learning ability of DRSN under different SNRs. Moreover, a joint mixed scenario training strategy is carried out to test the multi scenarios robustness of DRSN. As for the findings, the numerical results indicate that the DRSN method outperforms the spatial-frequency-temporal convolutional neural networks (SF-CNN)with similar computational complexity and achieves better advantages in the full SNR range than the minimum mean squared error (MMSE) estimator with a limited dataset. Moreover, the DRSN approach shows robustness in different propagation environments.展开更多
Recently,deep learning(DL)has been widely used in the field of remaining useful life(RUL)prediction.Among various DL technologies,recurrent neural network(RNN)and its variant,e.g.,long short-term memory(LSTM)network,h...Recently,deep learning(DL)has been widely used in the field of remaining useful life(RUL)prediction.Among various DL technologies,recurrent neural network(RNN)and its variant,e.g.,long short-term memory(LSTM)network,have gained extensive attention for their ability to capture temporal dependence.Although existing RNN-based methods have demonstrated their RUL prediction effectiveness,they still suffer from the following two limitations:1)it is difficult for the RNN to directly extract degradation features from original monitoring data and 2)most RNN-based prognostics methods are unable to quantify RUL uncertainty.To address the aforementioned limitations,this paper proposes a new prognostics method named residual convolution LSTM(RC-LSTM)network.In the RC-LSTM,a new ResNet-based convolution LSTM(Res-ConvLSTM)layer is stacked with a convolution LSTM(ConvLSTM)layer to extract degradation representations from monitoring data.Then,under the assumption that the RUL follows a normal distribution,an appropriate output layer is constructed to quantify the uncertainty of prediction results.Finally,the effectiveness and superiority of the RC-LSTM are verified using monitoring data from accelerated bearing degradation tests.展开更多
This study proposes a multi-scale simplified residual convolutional neural network(MS-SRCNN)for the precise prediction of Mg-Nd binary alloy compositions from scanning electron microscope(SEM)images.A multi-scale data...This study proposes a multi-scale simplified residual convolutional neural network(MS-SRCNN)for the precise prediction of Mg-Nd binary alloy compositions from scanning electron microscope(SEM)images.A multi-scale data structure is established by spatially aligning and stacking SEM images at different magnifications.The MS-SRCNN significantly reduces computational runtime by over 90%compared to traditional architectures like ResNet50,VGG16,and VGG19,without compromising prediction accuracy.The model demonstrates more excellent predictive performance,achieving a>5%increase in R^(2) compared to single-scale models.Furthermore,the MS-SRCNN exhibits robust composition prediction capability across other Mg-based binary alloys,including Mg-La,Mg-Sn,Mg-Ce,Mg-Sm,Mg-Ag,and Mg-Y,thereby emphasizing its generalization and extrapolation potential.This research establishes a non-destructive,microstructure-informed composition analysis framework,reduces characterization time compared to traditional experiment methods and provides insights into the composition-microstructure relationship in diverse material systems.展开更多
The enhanced variable rate codec (EVRC) is a standard for the 'Speech ServiceOption 3 for Wideband Spread Spectrum Digital System,' which has been employed in both IS-95cellular systems and ANSI J-STC-008 PCS ...The enhanced variable rate codec (EVRC) is a standard for the 'Speech ServiceOption 3 for Wideband Spread Spectrum Digital System,' which has been employed in both IS-95cellular systems and ANSI J-STC-008 PCS (personal communications systems). This paper concentrateson channel decoders that exploit the residual redundancy inherent in the enhanced variable ratecodec bitstream. This residual redundancy is quantified by modeling the parameters as first orderMarkov chains and computing the entropy rate based on the relative frequencies of transitions.Moreover, this residual redundancy can be exploited by an appropriately 'tuned' channel decoder toprovide substantial coding gain when compared with the decoders that do not exploit it. Channelcoding schemes include convolutional codes, and iteratively decoded parallel concatenatedconvolutional 'turbo' codes.展开更多
Taking the real part and the imaginary part of complex sound pressure of the sound field as features,a transfer learning model is constructed.Based on the pre-training of a large amount of underwater acoustic data in ...Taking the real part and the imaginary part of complex sound pressure of the sound field as features,a transfer learning model is constructed.Based on the pre-training of a large amount of underwater acoustic data in the preselected sea area using the convolutional neural network(CNN),the few-shot underwater acoustic data in the test sea area are retrained to study the underwater sound source ranging problem.The S5 voyage data of SWellEX-96 experiment is used to verify the proposed method,realize the range estimation for the shallow source in the experiment,and compare the range estimation performance of the underwater target sound source of four methods:matched field processing(MFP),generalized regression neural network(GRNN),traditional CNN,and transfer learning.Experimental data processing results show that the transfer learning model based on residual CNN can effectively realize range estimation in few-shot scenes,and the estimation performance is remarkably better than that of other methods.展开更多
Even though much advancements have been achieved with regards to the recognition of handwritten characters,researchers still face difficulties with the handwritten character recognition problem,especially with the adv...Even though much advancements have been achieved with regards to the recognition of handwritten characters,researchers still face difficulties with the handwritten character recognition problem,especially with the advent of new datasets like the Extended Modified National Institute of Standards and Technology dataset(EMNIST).The EMNIST dataset represents a challenge for both machine-learning and deep-learning techniques due to inter-class similarity and intra-class variability.Inter-class similarity exists because of the similarity between the shapes of certain characters in the dataset.The presence of intra-class variability is mainly due to different shapes written by different writers for the same character.In this research,we have optimized a deep residual network to achieve higher accuracy vs.the published state-of-the-art results.This approach is mainly based on the prebuilt deep residual network model ResNet18,whose architecture has been enhanced by using the optimal number of residual blocks and the optimal size of the receptive field of the first convolutional filter,the replacement of the first max-pooling filter by an average pooling filter,and the addition of a drop-out layer before the fully connected layer.A distinctive modification has been introduced by replacing the final addition layer with a depth concatenation layer,which resulted in a novel deep architecture having higher accuracy vs.the pure residual architecture.Moreover,the dataset images’sizes have been adjusted to optimize their visibility in the network.Finally,by tuning the training hyperparameters and using rotation and shear augmentations,the proposed model outperformed the state-of-the-art models by achieving average accuracies of 95.91%and 90.90%for the Letters and Balanced dataset sections,respectively.Furthermore,the average accuracies were improved to 95.9%and 91.06%for the Letters and Balanced sections,respectively,by using a group of 5 instances of the trained models and averaging the output class probabilities.展开更多
Automatic modulation recognition(AMR)of radiation source signals is a research focus in the field of cognitive radio.However,the AMR of radiation source signals at low SNRs still faces a great challenge.Therefore,the ...Automatic modulation recognition(AMR)of radiation source signals is a research focus in the field of cognitive radio.However,the AMR of radiation source signals at low SNRs still faces a great challenge.Therefore,the AMR method of radiation source signals based on two-dimensional data matrix and improved residual neural network is proposed in this paper.First,the time series of the radiation source signals are reconstructed into two-dimensional data matrix,which greatly simplifies the signal preprocessing process.Second,the depthwise convolution and large-size convolutional kernels based residual neural network(DLRNet)is proposed to improve the feature extraction capability of the AMR model.Finally,the model performs feature extraction and classification on the two-dimensional data matrix to obtain the recognition vector that represents the signal modulation type.Theoretical analysis and simulation results show that the AMR method based on two-dimensional data matrix and improved residual network can significantly improve the accuracy of the AMR method.The recognition accuracy of the proposed method maintains a high level greater than 90% even at -14 dB SNR.展开更多
Achieving accurate classification of colorectal polyps during colonoscopy can avoid unnecessary endoscopic biopsy or resection.This study aimed to develop a deep learning model that can automatically classify colorect...Achieving accurate classification of colorectal polyps during colonoscopy can avoid unnecessary endoscopic biopsy or resection.This study aimed to develop a deep learning model that can automatically classify colorectal polyps histologically on white-light and narrow-band imaging(NBI)colonoscopy images based on World Health Organization(WHO)and Workgroup serrAted polypS and Polyposis(WASP)classification criteria for colorectal polyps.White-light and NBI colonoscopy images of colorectal polyps exhibiting pathological results were firstly collected and classified into four categories:conventional adenoma,hyperplastic polyp,sessile serrated adenoma/polyp(SSAP)and normal,among which conventional adenoma could be further divided into three sub-categories of tubular adenoma,villous adenoma and villioustublar adenoma,subsequently the images were re-classified into six categories.In this paper,we proposed a novel convolutional neural network termed Polyp-DedNet for the four-and six-category classification tasks of colorectal polyps.Based on the existing classification network ResNet50,Polyp-DedNet adopted dilated convolution to retain more high-dimensional spatial information and an Efficient Channel Attention(ECA)module to improve the classification performance further.To eliminate gridding artifacts caused by dilated convolutions,traditional convolutional layers were used instead of the max pooling layer,and two convolutional layers with progressively decreasing dilation were added at the end of the network.Due to the inevitable imbalance of medical image data,a regularization method DropBlock and a Class-Balanced(CB)Loss were performed to prevent network overfitting.Furthermore,the 5-fold cross-validation was adopted to estimate the performance of Polyp-DedNet for the multi-classification task of colorectal polyps.Mean accuracies of the proposed Polyp-DedNet for the four-and six-category classifications of colorectal polyps were 89.91%±0.92%and 85.13%±1.10%,respectively.The metrics of precision,recall and F1-score were also improved by 1%∼2%compared to the baseline ResNet50.The proposed Polyp-DedNet presented state-of-the-art performance for colorectal polyp classifying on white-light and NBI colonoscopy images,highlighting its considerable potential as an AI-assistant system for accurate colorectal polyp diagnosis in colonoscopy.展开更多
基金supported by the Intelligent System Research Group(ISysRG)supported by Universitas Sriwijaya funded by the Competitive Research 2024.
文摘Handling missing data accurately is critical in clinical research, where data quality directly impacts decision-making and patient outcomes. While deep learning (DL) techniques for data imputation have gained attention, challenges remain, especially when dealing with diverse data types. In this study, we introduce a novel data imputation method based on a modified convolutional neural network, specifically, a Deep Residual-Convolutional Neural Network (DRes-CNN) architecture designed to handle missing values across various datasets. Our approach demonstrates substantial improvements over existing imputation techniques by leveraging residual connections and optimized convolutional layers to capture complex data patterns. We evaluated the model on publicly available datasets, including Medical Information Mart for Intensive Care (MIMIC-III and MIMIC-IV), which contain critical care patient data, and the Beijing Multi-Site Air Quality dataset, which measures environmental air quality. The proposed DRes-CNN method achieved a root mean square error (RMSE) of 0.00006, highlighting its high accuracy and robustness. We also compared with Low Light-Convolutional Neural Network (LL-CNN) and U-Net methods, which had RMSE values of 0.00075 and 0.00073, respectively. This represented an improvement of approximately 92% over LL-CNN and 91% over U-Net. The results showed that this DRes-CNN-based imputation method outperforms current state-of-the-art models. These results established DRes-CNN as a reliable solution for addressing missing data.
文摘In the burgeoning field of anomaly detection within attributed networks,traditional methodologies often encounter the intricacies of network complexity,particularly in capturing nonlinearity and sparsity.This study introduces an innovative approach that synergizes the strengths of graph convolutional networks with advanced deep residual learning and a unique residual-based attention mechanism,thereby creating a more nuanced and efficient method for anomaly detection in complex networks.The heart of our model lies in the integration of graph convolutional networks that capture complex structural relationships within the network data.This is further bolstered by deep residual learning,which is employed to model intricate nonlinear connections directly from input data.A pivotal innovation in our approach is the incorporation of a residual-based attention mech-anism.This mechanism dynamically adjusts the importance of nodes based on their residual information,thereby significantly enhancing the sensitivity of the model to subtle anomalies.Furthermore,we introduce a novel hypersphere mapping technique in the latent space to distinctly separate normal and anomalous data.This mapping is the key to our model’s ability to pinpoint anomalies with greater precision.An extensive experimental setup was used to validate the efficacy of the proposed model.Using attributed social network datasets,we demonstrate that our model not only competes with but also surpasses existing state-of-the-art methods in anomaly detection.The results show the exceptional capability of our model to handle the multifaceted nature of real-world networks.
基金the National Key Research and Development Program of China(No.2023 YFC2811600)the National Natural Science Foundation of China(Nos.52301349,52088102)+1 种基金the Major Science and Technology Innovation Program of Qingdao(No.223-3-hygg-10-hy)the Qingdao Science Foundation for Post-doctoral Scientists(Nos.QDBSH20220202070,QDBSH20220201015)。
文摘A deep-sea riser is a crucial component of the mining system used to lift seafloor mineral resources to the vessel.Even minor damage to the riser can lead to substantial financial losses,environmental impacts,and safety hazards.However,identifying modal parameters for structural health monitoring remains a major challenge due to its large deformations and flexibility.Vibration signal-based methods are essential for detecting damage and enabling timely maintenance to minimize losses.However,accurately extracting features from one-dimensional(1D)signals is often hindered by various environmental factors and measurement noises.To address this challenge,a novel approach based on a residual convolutional auto-encoder(RCAE)is proposed for detecting damage in deep-sea mining risers,incorporating a data fusion strategy.First,principal component analysis(PCA)is applied to reduce environmental fluctuations and fuse multisensor strain readings.Subsequently,a 1D-RCAE is used to extract damage-sensitive features(DSFs)from the fused dataset.A Mahalanobis distance indicator is established to compare the DSFs of the testing and healthy risers.The specific threshold for these distances is determined using the 3σcriterion,which is employed to assess whether damage has occurred in the testing riser.The effectiveness and robustness of the proposed approach are verified through numerical simulations of a 500-m riser and experimental tests on a 6-m riser.Moreover,the impact of contaminated noise and environmental fluctuations is examined.Results show that the proposed PCA-1D-RCAE approach can effectively detect damage and is resilient to measurement noise and environmental fluctuations.The accuracy exceeds 98%under noise-free conditions and remains above 90%even with 10 dB noise.This novel approach has the potential to establish a new standard for evaluating the health and integrity of risers during mining operations,thereby reducing the high costs and risks associated with failures.Maintenance activities can be scheduled more efficiently by enabling early and accurate detection of riser damage,minimizing downtime and avoiding catastrophic failures.
文摘In daily life,keyword spotting plays an important role in human-computer interaction.However,noise often interferes with the extraction of time-frequency information,and achieving both computational efficiency and recognition accuracy on resource-constrained devices such as mobile terminals remains a major challenge.To address this,we propose a novel time-frequency dual-branch parallel residual network,which integrates a Dual-Branch Broadcast Residual module and a Time-Frequency Coordinate Attention module.The time-domain and frequency-domain branches are designed in parallel to independently extract temporal and spectral features,effectively avoiding the potential information loss caused by serial stacking,while enhancing information flow and multi-scale feature fusion.In terms of training strategy,a curriculum learning approach is introduced to progressively improve model robustness fromeasy to difficult tasks.Experimental results demonstrate that the proposed method consistently outperforms existing lightweight models under various signal-to-noise ratio(SNR)conditions,achieving superior far-field recognition performance on the Google Speech Commands V2 dataset.Notably,the model maintains stable performance even in low-SNR environments such as–10 dB,and generalizes well to unseen SNR conditions during training,validating its robustness to novel noise scenarios.Furthermore,the proposed model exhibits significantly fewer parameters,making it highly suitable for deployment on resource-limited devices.Overall,the model achieves a favorable balance between performance and parameter efficiency,demonstrating strong potential for practical applications.
基金supported by the Xiamen Science and Technology Subsidy Project(No.2023CXY0318).
文摘Abnormal network traffic, as a frequent security risk, requires a series of techniques to categorize and detect it. Existing network traffic anomaly detection still faces challenges: the inability to fully extract local and global features, as well as the lack of effective mechanisms to capture complex interactions between features;Additionally, when increasing the receptive field to obtain deeper feature representations, the reliance on increasing network depth leads to a significant increase in computational resource consumption, affecting the efficiency and performance of detection. Based on these issues, firstly, this paper proposes a network traffic anomaly detection model based on parallel dilated convolution and residual learning (Res-PDC). To better explore the interactive relationships between features, the traffic samples are converted into two-dimensional matrix. A module combining parallel dilated convolutions and residual learning (res-pdc) was designed to extract local and global features of traffic at different scales. By utilizing res-pdc modules with different dilation rates, we can effectively capture spatial features at different scales and explore feature dependencies spanning wider regions without increasing computational resources. Secondly, to focus and integrate the information in different feature subspaces, further enhance and extract the interactions among the features, multi-head attention is added to Res-PDC, resulting in the final model: multi-head attention enhanced parallel dilated convolution and residual learning (MHA-Res-PDC) for network traffic anomaly detection. Finally, comparisons with other machine learning and deep learning algorithms are conducted on the NSL-KDD and CIC-IDS-2018 datasets. The experimental results demonstrate that the proposed method in this paper can effectively improve the detection performance.
基金supported by Nature Science Research Project of Higher Education Institutions in Jiangsu Province under Grant No.21KJB510018National Nature Science Foundation of China (NSFC)under Grant No.62001215.
文摘Microphone array-based sound source localization(SSL)is widely used in a variety of occasions such as video conferencing,robotic hearing,speech enhancement,speech recognition and so on.The traditional SSL methods cannot achieve satisfactory performance in adverse noisy and reverberant environments.In order to improve localization performance,a novel SSL algorithm using convolutional residual network(CRN)is proposed in this paper.The spatial features including time difference of arrivals(TDOAs)between microphone pairs and steered response power-phase transform(SRPPHAT)spatial spectrum are extracted in each Gammatone sub-band.The spatial features of different sub-bands with a frame are combine into a feature matrix as the input of CRN.The proposed algorithm employ CRN to fuse the spatial features.Since the CRN introduces the residual structure on the basis of the convolutional network,it reduce the difficulty of training procedure and accelerate the convergence of the model.A CRN model is learned from the training data in various reverberation and noise environments to establish the mapping regularity between the input feature and the sound azimuth.Through simulation verification,compared with the methods using traditional deep neural network,the proposed algorithm can achieve a better localization performance in SSL task,and provide better generalization capacity to untrained noise and reverberation.
基金Fundamental Research Funds for the Central Universities(Grant No.FRF-TP-19-006A3).
文摘As a common and high-risk type of disease,heart disease seriously threatens people’s health.At the same time,in the era of the Internet of Thing(IoT),smart medical device has strong practical significance for medical workers and patients because of its ability to assist in the diagnosis of diseases.Therefore,the research of real-time diagnosis and classification algorithms for arrhythmia can help to improve the diagnostic efficiency of diseases.In this paper,we design an automatic arrhythmia classification algorithm model based on Convolutional Neural Network(CNN)and Encoder-Decoder model.The model uses Long Short-Term Memory(LSTM)to consider the influence of time series features on classification results.Simultaneously,it is trained and tested by the MIT-BIH arrhythmia database.Besides,Generative Adversarial Networks(GAN)is adopted as a method of data equalization for solving data imbalance problem.The simulation results show that for the inter-patient arrhythmia classification,the hybrid model combining CNN and Encoder-Decoder model has the best classification accuracy,of which the accuracy can reach 94.05%.Especially,it has a better advantage for the classification effect of supraventricular ectopic beats(class S)and fusion beats(class F).
基金supported by the National Natural Science Foundation of China(No.41906169)the PLA Academy of Military Sciences.
文摘Noise reduction analysis of signals is essential for modern underwater acoustic detection systems.The traditional noise reduction techniques gradually lose efficacy because the target signal is masked by biological and natural noise in the marine environ-ment.The feature extraction method combining time-frequency spectrograms and deep learning can effectively achieve the separation of noise and target signals.A fully convolutional encoder-decoder neural network(FCEDN)is proposed to address the issue of noise reduc-tion in underwater acoustic signals.The time-domain waveform map of underwater acoustic signals is converted into a wavelet low-frequency analysis recording spectrogram during the denoising process to preserve as many underwater acoustic signal characteristics as possible.The FCEDN is built to learn the spectrogram mapping between noise and target signals that can be learned at each time level.The transposed convolution transforms are introduced,which can transform the spectrogram features of the signals into listenable audio files.After evaluating the systems on the ShipsEar Dataset,the proposed method can increase SNR and SI-SNR by 10.02 and 9.5dB,re-spectively.
基金This work was supported in part by the science and technology research project of Henan Provincial Department of science and technology(No.222102110366)the Science and Technology Innovation Team of Henan University(No.22IRTSTHN016)the grants from the teaching reform research and practice project of higher education in Henan Province in 2021[2021SJGLX502].
文摘Amodel that can obtain rapid and accurate detection of coronavirus disease 2019(COVID-19)plays a significant role in treating and preventing the spread of disease transmission.However,designing such amodel that can balance the detection accuracy andweight parameters ofmemorywell to deploy a mobile device is challenging.Taking this point into account,this paper fuses the convolutional neural network and residual learning operations to build a multi-class classification model,which improves COVID-19 pneumonia detection performance and keeps a trade-off between the weight parameters and accuracy.The convolutional neural network can extract the COVID-19 feature information by repeated convolutional operations.The residual learning operations alleviate the gradient problems caused by stacking convolutional layers and enhance the ability of feature extraction.The ability further enables the proposed model to acquire effective feature information at a lowcost,which canmake ourmodel keep smallweight parameters.Extensive validation and comparison with other models of COVID-19 pneumonia detection on the well-known COVIDx dataset show that(1)the sensitivity of COVID-19 pneumonia detection is improved from 88.2%(non-COVID-19)and 77.5%(COVID-19)to 95.3%(non-COVID-19)and 96.5%(COVID-19),respectively.The positive predictive value is also respectively increased from72.8%(non-COVID-19)and 89.0%(COVID-19)to 88.8%(non-COVID-19)and 95.1%(COVID-19).(2)Compared with the weight parameters of the COVIDNet-small network,the value of the proposed model is 13 M,which is slightly higher than that(11.37 M)of the COVIDNet-small network.But,the corresponding accuracy is improved from 85.2%to 93.0%.The above results illustrate the proposed model can gain an efficient balance between accuracy and weight parameters.
基金National Natural Science Foundation of China(No.52101346)Fundamental Research Funds for the Central Universities,China(No.2232019D3-61)Initial Research Fund for the Young Teachers of Donghua University,China。
文摘Robotic grasps play an important role in the service and industrial fields,and the robotic arm can grasp the object properly depends on the accuracy of the grasping detection result.In order to predict grasping detection positions for known or unknown objects by a modular robotic system,a convolutional neural network(CNN)with the residual block is proposed,which can be used to generate accurate grasping detection for input images of the scene.The proposed model architecture was trained on the standard Cornell grasp dataset and evaluated on the test dataset.Moreover,it was evaluated on different types of household objects and cluttered multi-objects.On the Cornell grasp dataset,the accuracy of the model on image-wise splitting detection and object-wise splitting detection achieved 95.5%and 93.6%,respectively.Further,the real detection time per image was 109 ms.The experimental results show that the model can quickly detect the grasping positions of a single object or multiple objects in image pixels in real time,and it keeps good stability and robustness.
基金supported in part by the Basic and Applied Basic Research Foundation of Guangdong Province[2025A1515011566]in part by the State Key Laboratory for Novel Software Technology,Nanjing University[KFKT2024B08]+1 种基金in part by Leading Talents in Gusu Innovation and Entrepreneurship[ZXL2023170]in part by the Basic Research Programs of Taicang 2024,[TC2024JC32].
文摘Deep convolutional neural networks(CNNs)have demonstrated remarkable performance in video super-resolution(VSR).However,the ability of most existing methods to recover fine details in complex scenes is often hindered by the loss of shallow texture information during feature extraction.To address this limitation,we propose a 3D Convolutional Enhanced Residual Video Super-Resolution Network(3D-ERVSNet).This network employs a forward and backward bidirectional propagation module(FBBPM)that aligns features across frames using explicit optical flow through lightweight SPyNet.By incorporating an enhanced residual structure(ERS)with skip connections,shallow and deep features are effectively integrated,enhancing texture restoration capabilities.Furthermore,3D convolution module(3DCM)is applied after the backward propagation module to implicitly capture spatio-temporal dependencies.The architecture synergizes these components where FBBPM extracts aligned features,ERS fuses hierarchical representations,and 3DCM refines temporal coherence.Finally,a deep feature aggregation module(DFAM)fuses the processed features,and a pixel-upsampling module(PUM)reconstructs the high-resolution(HR)video frames.Comprehensive evaluations on REDS,Vid4,UDM10,and Vim4 benchmarks demonstrate well performance including 30.95 dB PSNR/0.8822 SSIM on REDS and 32.78 dB/0.8987 on Vim4.3D-ERVSNet achieves significant gains over baselines while maintaining high efficiency with only 6.3M parameters and 77ms/frame runtime(i.e.,20×faster than RBPN).The network’s effectiveness stems from its task-specific asymmetric design that balances explicit alignment and implicit fusion.
基金supported by the National Key Scientific Instrument and Equipment Development Project(61827801).
文摘The fifth-generation (5G) communication requires a highly accurate estimation of the channel state information (CSI)to take advantage of the massive multiple-input multiple-output(MIMO) system. However, traditional channel estimation methods do not always yield reliable estimates. The methodology of this paper consists of deep residual shrinkage network (DRSN)neural network-based method that is used to solve this problem.Thus, the channel estimation approach, based on DRSN with its learning ability of noise-containing data, is first introduced. Then,the DRSN is used to train the noise reduction process based on the results of the least square (LS) channel estimation while applying the pilot frequency subcarriers, where the initially estimated subcarrier channel matrix is considered as a three-dimensional tensor of the DRSN input. Afterward, a mixed signal to noise ratio (SNR) training data strategy is proposed based on the learning ability of DRSN under different SNRs. Moreover, a joint mixed scenario training strategy is carried out to test the multi scenarios robustness of DRSN. As for the findings, the numerical results indicate that the DRSN method outperforms the spatial-frequency-temporal convolutional neural networks (SF-CNN)with similar computational complexity and achieves better advantages in the full SNR range than the minimum mean squared error (MMSE) estimator with a limited dataset. Moreover, the DRSN approach shows robustness in different propagation environments.
基金This research was supported by National Natural Science Foundation of China(52005387,52025056)Project funded by China Postdoctoral Science Foundation(2020M673380)Fundamental Research Funds for the Central Universities.
文摘Recently,deep learning(DL)has been widely used in the field of remaining useful life(RUL)prediction.Among various DL technologies,recurrent neural network(RNN)and its variant,e.g.,long short-term memory(LSTM)network,have gained extensive attention for their ability to capture temporal dependence.Although existing RNN-based methods have demonstrated their RUL prediction effectiveness,they still suffer from the following two limitations:1)it is difficult for the RNN to directly extract degradation features from original monitoring data and 2)most RNN-based prognostics methods are unable to quantify RUL uncertainty.To address the aforementioned limitations,this paper proposes a new prognostics method named residual convolution LSTM(RC-LSTM)network.In the RC-LSTM,a new ResNet-based convolution LSTM(Res-ConvLSTM)layer is stacked with a convolution LSTM(ConvLSTM)layer to extract degradation representations from monitoring data.Then,under the assumption that the RUL follows a normal distribution,an appropriate output layer is constructed to quantify the uncertainty of prediction results.Finally,the effectiveness and superiority of the RC-LSTM are verified using monitoring data from accelerated bearing degradation tests.
基金funded by the National Natural Science Foundation of China(No.52204407)the Natural Science Foundation of Jiangsu Province(No.BK20220595)the China Postdoctoral Science Foundation(No.2022M723689).
文摘This study proposes a multi-scale simplified residual convolutional neural network(MS-SRCNN)for the precise prediction of Mg-Nd binary alloy compositions from scanning electron microscope(SEM)images.A multi-scale data structure is established by spatially aligning and stacking SEM images at different magnifications.The MS-SRCNN significantly reduces computational runtime by over 90%compared to traditional architectures like ResNet50,VGG16,and VGG19,without compromising prediction accuracy.The model demonstrates more excellent predictive performance,achieving a>5%increase in R^(2) compared to single-scale models.Furthermore,the MS-SRCNN exhibits robust composition prediction capability across other Mg-based binary alloys,including Mg-La,Mg-Sn,Mg-Ce,Mg-Sm,Mg-Ag,and Mg-Y,thereby emphasizing its generalization and extrapolation potential.This research establishes a non-destructive,microstructure-informed composition analysis framework,reduces characterization time compared to traditional experiment methods and provides insights into the composition-microstructure relationship in diverse material systems.
基金supported by the Fundamental Research Funds for the Central Universities (No.2022JCCXMT01)the National College Students’Innovation and Entrepreneurship Training Program Automatic Recognition of Earthquake Faults Based on Convolutional Neural Networks (No.20220236)the Open Fund of State Key Laboratory for Fine Exploration and Intelligent Development of Coal Resources (No.SKLCRSM22DC02).
文摘The enhanced variable rate codec (EVRC) is a standard for the 'Speech ServiceOption 3 for Wideband Spread Spectrum Digital System,' which has been employed in both IS-95cellular systems and ANSI J-STC-008 PCS (personal communications systems). This paper concentrateson channel decoders that exploit the residual redundancy inherent in the enhanced variable ratecodec bitstream. This residual redundancy is quantified by modeling the parameters as first orderMarkov chains and computing the entropy rate based on the relative frequencies of transitions.Moreover, this residual redundancy can be exploited by an appropriately 'tuned' channel decoder toprovide substantial coding gain when compared with the decoders that do not exploit it. Channelcoding schemes include convolutional codes, and iteratively decoded parallel concatenatedconvolutional 'turbo' codes.
基金supported by the National Natural Science Foundation of China(1197428611904274)+1 种基金the Shaanxi Young Science and Technology Star Program(2021KJXX-07)the fundamental research funding for characteristic disciplines(G2022WD0235)。
文摘Taking the real part and the imaginary part of complex sound pressure of the sound field as features,a transfer learning model is constructed.Based on the pre-training of a large amount of underwater acoustic data in the preselected sea area using the convolutional neural network(CNN),the few-shot underwater acoustic data in the test sea area are retrained to study the underwater sound source ranging problem.The S5 voyage data of SWellEX-96 experiment is used to verify the proposed method,realize the range estimation for the shallow source in the experiment,and compare the range estimation performance of the underwater target sound source of four methods:matched field processing(MFP),generalized regression neural network(GRNN),traditional CNN,and transfer learning.Experimental data processing results show that the transfer learning model based on residual CNN can effectively realize range estimation in few-shot scenes,and the estimation performance is remarkably better than that of other methods.
文摘Even though much advancements have been achieved with regards to the recognition of handwritten characters,researchers still face difficulties with the handwritten character recognition problem,especially with the advent of new datasets like the Extended Modified National Institute of Standards and Technology dataset(EMNIST).The EMNIST dataset represents a challenge for both machine-learning and deep-learning techniques due to inter-class similarity and intra-class variability.Inter-class similarity exists because of the similarity between the shapes of certain characters in the dataset.The presence of intra-class variability is mainly due to different shapes written by different writers for the same character.In this research,we have optimized a deep residual network to achieve higher accuracy vs.the published state-of-the-art results.This approach is mainly based on the prebuilt deep residual network model ResNet18,whose architecture has been enhanced by using the optimal number of residual blocks and the optimal size of the receptive field of the first convolutional filter,the replacement of the first max-pooling filter by an average pooling filter,and the addition of a drop-out layer before the fully connected layer.A distinctive modification has been introduced by replacing the final addition layer with a depth concatenation layer,which resulted in a novel deep architecture having higher accuracy vs.the pure residual architecture.Moreover,the dataset images’sizes have been adjusted to optimize their visibility in the network.Finally,by tuning the training hyperparameters and using rotation and shear augmentations,the proposed model outperformed the state-of-the-art models by achieving average accuracies of 95.91%and 90.90%for the Letters and Balanced dataset sections,respectively.Furthermore,the average accuracies were improved to 95.9%and 91.06%for the Letters and Balanced sections,respectively,by using a group of 5 instances of the trained models and averaging the output class probabilities.
基金National Natural Science Foundation of China under Grant No.61973037China Postdoctoral Science Foundation under Grant No.2022M720419。
文摘Automatic modulation recognition(AMR)of radiation source signals is a research focus in the field of cognitive radio.However,the AMR of radiation source signals at low SNRs still faces a great challenge.Therefore,the AMR method of radiation source signals based on two-dimensional data matrix and improved residual neural network is proposed in this paper.First,the time series of the radiation source signals are reconstructed into two-dimensional data matrix,which greatly simplifies the signal preprocessing process.Second,the depthwise convolution and large-size convolutional kernels based residual neural network(DLRNet)is proposed to improve the feature extraction capability of the AMR model.Finally,the model performs feature extraction and classification on the two-dimensional data matrix to obtain the recognition vector that represents the signal modulation type.Theoretical analysis and simulation results show that the AMR method based on two-dimensional data matrix and improved residual network can significantly improve the accuracy of the AMR method.The recognition accuracy of the proposed method maintains a high level greater than 90% even at -14 dB SNR.
基金funded by the Research Fund for Foundation of Hebei University(DXK201914)the President of Hebei University(XZJJ201914)+3 种基金the Post-graduate’s Innovation Fund Project of Hebei University(HBU2022SS003)the Special Project for Cultivating College Students’Scientific and Technological Innovation Ability in Hebei Province(22E50041D)Guangdong Basic and Applied Basic Research Foundation(2021A1515011654)the Fundamental Research Funds for the Central Universities of China(20720210117).
文摘Achieving accurate classification of colorectal polyps during colonoscopy can avoid unnecessary endoscopic biopsy or resection.This study aimed to develop a deep learning model that can automatically classify colorectal polyps histologically on white-light and narrow-band imaging(NBI)colonoscopy images based on World Health Organization(WHO)and Workgroup serrAted polypS and Polyposis(WASP)classification criteria for colorectal polyps.White-light and NBI colonoscopy images of colorectal polyps exhibiting pathological results were firstly collected and classified into four categories:conventional adenoma,hyperplastic polyp,sessile serrated adenoma/polyp(SSAP)and normal,among which conventional adenoma could be further divided into three sub-categories of tubular adenoma,villous adenoma and villioustublar adenoma,subsequently the images were re-classified into six categories.In this paper,we proposed a novel convolutional neural network termed Polyp-DedNet for the four-and six-category classification tasks of colorectal polyps.Based on the existing classification network ResNet50,Polyp-DedNet adopted dilated convolution to retain more high-dimensional spatial information and an Efficient Channel Attention(ECA)module to improve the classification performance further.To eliminate gridding artifacts caused by dilated convolutions,traditional convolutional layers were used instead of the max pooling layer,and two convolutional layers with progressively decreasing dilation were added at the end of the network.Due to the inevitable imbalance of medical image data,a regularization method DropBlock and a Class-Balanced(CB)Loss were performed to prevent network overfitting.Furthermore,the 5-fold cross-validation was adopted to estimate the performance of Polyp-DedNet for the multi-classification task of colorectal polyps.Mean accuracies of the proposed Polyp-DedNet for the four-and six-category classifications of colorectal polyps were 89.91%±0.92%and 85.13%±1.10%,respectively.The metrics of precision,recall and F1-score were also improved by 1%∼2%compared to the baseline ResNet50.The proposed Polyp-DedNet presented state-of-the-art performance for colorectal polyp classifying on white-light and NBI colonoscopy images,highlighting its considerable potential as an AI-assistant system for accurate colorectal polyp diagnosis in colonoscopy.
