Convolutional Classifier Research Articles

Efficacious human metapneumovirus vaccine based on AI-guided engineering of a closed prefusion trimer

The prefusion conformation of human metapneumovirus fusion protein (hMPV Pre-F) is critical for eliciting the most potent neutralizing antibodies and is the preferred immunogen for an efficacious vaccine against hMPV respiratory infections. Here we show that an additional cleavage event in the F protein allows closure and correct folding of the trimer. We therefore engineered the F protein to undergo double cleavage, which enabled screening for Pre-F stabilizing substitutions at the natively folded protomer interfaces. To identify these substitutions, we developed an AI convolutional classifier that successfully predicts complex polar interactions often overlooked by physics-based methods and visual inspection. The combination of additional processing, stabilization of interface regions and stabilization of the membrane-proximal stem, resulted in a Pre-F protein vaccine candidate without the need for a heterologous trimerization domain that exhibited high expression yields and thermostability. Cryo-EM analysis shows the complete ectodomain structure, including the stem, and a specific interaction of the newly identified cleaved C-terminus with the adjacent protomer. Importantly, the protein induces high and cross-neutralizing antibody responses resulting in near complete protection against hMPV challenge in cotton rats, making the highly stable, double-cleaved hMPV Pre-F trimer an attractive vaccine candidate.

Detection and mitigation of DDoS attacks based on multi-dimensional characteristics in SDN

Due to the large computational overhead, underutilization of features, and high bandwidth consumption in traditional SDN environments for DDoS attack detection and mitigation methods, this paper proposes a two-stage detection and mitigation method for DDoS attacks in SDN based on multi-dimensional characteristics. Firstly, an analysis of the traffic statistics from the SDN switch ports is performed, which aids in conducting a coarse-grained detection of DDoS attacks within the network. Subsequently, a Multi-Dimensional Deep Convolutional Classifier (MDDCC) is constructed using wavelet decomposition and convolutional neural networks to extract multi-dimensional characteristics from the traffic data passing through suspicious switches. Based on these extracted multi-dimensional characteristics, a simple classifier can be employed to accurately detect attack samples. Finally, by integrating graph theory with restrictive strategies, the source of attacks in SDN networks can be effectively traced and isolated. The experimental results indicate that the proposed method, which utilizes a minimal amount of statistical information, can quickly and accurately detect attacks within the SDN network. It demonstrates superior accuracy and generalization capabilities compared to traditional detection methods, especially when tested on both simulated and public datasets. Furthermore, by isolating the affected nodes, the method effectively mitigates the impact of the attacks, ensuring the normal transmission of legitimate traffic during network attacks. This approach not only enhances the detection capabilities but also provides a robust mechanism for containing the spread of cyber threats, thereby safeguarding the integrity and performance of the network.

Data augmentation based on highlight image models of underwater maneuvering target

With the development of underwater acoustic countermeasure technology, deep learning is applied to recognize echo geometry features of underwater targets, but it faces the problem of sample scarcity. In this paper, we improved the underwater target highlight model, and established the target echo information equation of active sonar. By changing the spatial positions of target and sonar regularly, we performed the highlight image models of underwater maneuvering targets. Taking an underwater vehicle as an example, the model construction process was introduced in detail, and highlight image models of four typical acoustic scale decoys were also established, and five multi-space state highlight image data samples were generated. The eHasNet-5 convolutional classification network was designed, and the network was trained, verified and tested with the generated data. Finally, the experimental data test shows that the target highlight image generation models provide a new data augmentation method for the application of deep learning in active sonar target recognition, and the trained network by generated data has the ability to classify two-dimensional objects.

Leveraging Data Locality in Quantum Convolutional Classifiers.

Quantum computing (QC) has opened the door to advancements in machine learning (ML) tasks that are currently implemented in the classical domain. Convolutional neural networks (CNNs) are classical ML architectures that exploit data locality and possess a simpler structure than a fully connected multi-layer perceptrons (MLPs) without compromising the accuracy of classification. However, the concept of preserving data locality is usually overlooked in the existing quantum counterparts of CNNs, particularly for extracting multifeatures in multidimensional data. In this paper, we present an multidimensional quantum convolutional classifier (MQCC) that performs multidimensional and multifeature quantum convolution with average and Euclidean pooling, thus adapting the CNN structure to a variational quantum algorithm (VQA). The experimental work was conducted using multidimensional data to validate the correctness and demonstrate the scalability of the proposed method utilizing both noisy and noise-free quantum simulations. We evaluated the MQCC model with reference to reported work on state-of-the-art quantum simulators from IBM Quantum and Xanadu using a variety of standard ML datasets. The experimental results show the favorable characteristics of our proposed techniques compared with existing work with respect to a number of quantitative metrics, such as the number of training parameters, cross-entropy loss, classification accuracy, circuit depth, and quantum gate count.

Medical image foundation models in assisting diagnosis of brain tumors: a pilot study.

To build self-supervised foundation models for multicontrast MRI of the whole brain and evaluate their efficacy in assisting diagnosis of brain tumors. In this retrospective study, foundation models were developed using 57,621 enhanced head MRI scans through self-supervised learning with a pretext task of cross-contrast context restoration with two different content dropout schemes. Downstream classifiers were constructed based on the pretrained foundation models and fine-tuned for brain tumor detection, discrimination, and molecular status prediction. Metrics including accuracy, sensitivity, specificity, and area under the ROC curve (AUC) were used to evaluate the performance. Convolutional neural networks trained exclusively on downstream task data were employed for comparative analysis. The pretrained foundation models demonstrated their ability to extract effective representations from multicontrast whole-brain volumes. The best classifiers, endowed with pretrained weights, showed remarkable performance with accuracies of 94.9, 92.3, and 80.4%, and corresponding AUC values of 0.981, 0.972, and 0.852 on independent test datasets in brain tumor detection, discrimination, and molecular status prediction, respectively. The classifiers with pretrained weights outperformed the convolutional classifiers trained from scratch by approximately 10% in terms of accuracy and AUC across all tasks. The saliency regions in the correctly predicted cases are mainly clustered around the tumors. Classifiers derived from the two dropout schemes differed significantly only in the detection of brain tumors. Foundation models obtained from self-supervised learning have demonstrated encouraging potential for scalability and interpretability in downstream brain tumor-related tasks and hold promise for extension to neurological diseases with diffusely distributed lesions. The application of our proposed method to the prediction of key molecular status in gliomas is expected to improve treatment planning and patient outcomes. Additionally, the foundation model we developed could serve as a cornerstone for advancing AI applications in the diagnosis of brain-related diseases.

Compressing Image-to-Image Translation GANs Using Local Density Structures on Their Learned Manifold

Generative Adversarial Networks (GANs) have shown remarkable success in modeling complex data distributions for image-to-image translation. Still, their high computational demands prohibit their deployment in practical scenarios like edge devices. Existing GAN compression methods mainly rely on knowledge distillation or convolutional classifiers' pruning techniques. Thus, they neglect the critical characteristic of GANs: their local density structure over their learned manifold. Accordingly, we approach GAN compression from a new perspective by explicitly encouraging the pruned model to preserve the density structure of the original parameter-heavy model on its learned manifold. We facilitate this objective for the pruned model by partitioning the learned manifold of the original generator into local neighborhoods around its generated samples. Then, we propose a novel pruning objective to regularize the pruned model to preserve the local density structure over each neighborhood, resembling the kernel density estimation method. Also, we develop a collaborative pruning scheme in which the discriminator and generator are pruned by two pruning agents. We design the agents to capture interactions between the generator and discriminator by exchanging their peer's feedback when determining corresponding models' architectures. Thanks to such a design, our pruning method can efficiently find performant sub-networks and can maintain the balance between the generator and discriminator more effectively compared to baselines during pruning, thereby showing more stable pruning dynamics. Our experiments on image translation GAN models, Pix2Pix and CycleGAN, with various benchmark datasets and architectures demonstrate our method's effectiveness.

Electrocardiogram identification based on data generative network and non-fiducial data processing

Nowadays, the use of biological signals as a criterion for identity recognition has gained increasing attention from various organizations and companies. Therefore, it has become crucial to have a biometric identity recognition method that is fast and accurate. In this paper, we propose a linear electrocardiogram (ECG) data preprocessing algorithm based on Kalman filters for rapid noise data filtering (wavelet transform filtering algorithm). Additionally, we introduce a generative network model called Data Generation Strategy Network (DRCN) based on generative networks. The DRCN is employed to augment training samples for convolutional classification networks, ultimately improving the classification performance of the model. Through the final experiments, our method successfully reduced the average misidentification rate of ECG-based identity recognition to 2.5%, and achieved an average recognition rate of 98.7% for each category, significantly surpassing previous achievements. In the future, this method is expected to be widely applied in the field of ECG-based identity recognition.

Transforming two-dimensional tensor networks into quantum circuits for supervised learning

There have been numerous quantum neural networks reported, but they struggle to match traditional neural networks in accuracy. Given the huge improvement of the neural network models’ accuracy by two-dimensional tensor network (TN) states in classical tensor network machine learning (TNML), it is promising to explore whether its application in quantum machine learning can extend the performance boundary of the models. Here, we transform two-dimensional TNs into quantum circuits for supervised learning. Specifically, we encode two-dimensional TNs into quantum circuits through rigorous mathematical proofs for constructing model ansätze, including string-bond states, entangled-plaquette states and isometric TN states. In addition, we propose adaptive data encoding methods and combine with TNs. We construct a tensor-network-inspired quantum circuit (TNQC) supervised learning framework for transferring TNML from classical to quantum, and build several novel two-dimensional TN-inspired quantum classifiers based on this framework. Finally, we propose a parallel quantum machine learning method for multi-class classification to construct 2D TNQC-based multi-class classifiers. Classical simulation results on the MNIST benchmark dataset show that our proposed models achieve the state-of-the-art accuracy performance, significantly outperforming other quantum classifiers on both binary and multi-class classification tasks, and beat simple convolutional classifiers on a fair track with identical inputs. The noise resilience of the models makes them successfully run and work in a real quantum computer.

EEG-FRM: a neural network based familiar and unfamiliar face EEG recognition method.

Recognizing familiar faces holds great value in various fields such as medicine, criminal investigation, and lie detection. In this paper, we designed a Complex Trial Protocol-based familiar and unfamiliar face recognition experiment that using self-face information, and collected EEG data from 147 subjects. A novel neural network-based method, the EEG-based Face Recognition Model (EEG-FRM), is proposed in this paper for cross-subject familiar/unfamiliar face recognition, which combines a multi-scale convolutional classification network with the maximum probability mechanism to realize individual face recognition. The multi-scale convolutional neural network extracts temporal information and spatial features from the EEG data, the attention module and supervised contrastive learning module are employed to promote the classification performance. Experimental results on the dataset reveal that familiar face stimuli could evoke significant P300 responses, mainly concentrated in the parietal lobe and nearby regions. Our proposed model achieved impressive results, with a balanced accuracy of 85.64%, a true positive rate of 73.23%, and a false positive rate of 1.96% on the collected dataset, outperforming other compared methods. The experimental results demonstrate the effectiveness and superiority of our proposed model.

PCCN-MSS: Parallel Convolutional Classification Network Combined Multi-Spatial Scale and Spectral Features for UAV-Borne Hyperspectral With High Spatial Resolution Imagery

PCCN-MSS: Parallel Convolutional Classification Network Combined Multi-Spatial Scale and Spectral Features for UAV-Borne Hyperspectral With High Spatial Resolution Imagery

Text classification based on PEGCN: Graph convolution classification using location information and edge features

AbstractThe purpose of text classification is to label the text with known labels. In recent years, the method based on graph neural network (GNN) has achieved good results. However, the existing methods based on GNN only regard the text as the set of co‐occurring words, without considering the position information of each word in the statement. At the same time, the method mainly extracts the node features in the graph, and the edge features between the nodes are not used enough. To solve these problems, a new text classification method, graph convolutional network using positions and edges, is proposed. In the word embedding section, a positional encoding input representation is employed to enable the neural network to learn the relative positional information among words. Meanwhile, the dimension of the adjacency matrix is increased to extract the multi‐dimensional edge features. Through experiments on multiple text classification datasets, the proposed method is shown to be superior to the traditional text classification method, and has achieved a maximum improvement of more than 4%.

Liver tumour detection and classification using partial differential technique algorithm with enhanced convolutional classifier

The image of liver which is the area of interest in this work is obtained from abdominal CT scan. It also contains details of other abdominal organs such as pancreas, spleen, stomach, gall bladder, intestine etc. Since all these organs are of soft tissues, the pixel intensity values differ marginally in the CT scan output and the organs overlap each other at their boundaries. Hence it is very difficult to trace out the exact contour of liver and liver tumor. The overlapping and obscure boundaries are to be avoided for proper diagnosis. Image segmentation process helps to meet this requirement. The normal perception of the CT image can be improved by suitable segmentation techniques. This will help the physician to extract more information from the image and give an accurate diagnosis and better treatment. The projected images are processed using the Partial Differential Technique (PDT) to isolate the liver from the other organs. The Level Set Methodology (LSM) is then used to separate the cancerous tissue from the healthy tissue around it. The classification of stages may be done with the assistance of an Enhanced Convolutional Classifier. The classification of LSM is evaluated by producing many metrics of accuracy, sensitivity, and specificity using an Improved Convolutional classifier. Compared to the two current algorithms, the proposed technique has a sensitivity and specificity of 96% and 93%, respectively, with 95% confidence intervals of [0.7513 1.0000] and [0.7126 1.0000] for sensitivity, and specificity respectively.

Research on Financial Fraud Detection Models Integrating Multiple Relational Graphs

The current fraud risk in digital finance is increasing year by year, and the mainstream solutions rely on the inherent characteristics of users, which makes it difficult to explain fraud behaviors and fraud behavior patterns are less researched. To address these problems, we propose an integrated multiple relational graphs fraud detection model Tri-RGCN-XGBoost, which analyzes the impact of user association patterns on fraud detection by mining the behavioral associations of users. The model builds a heterogeneous information network based on real transaction data, abstracts three types of bipartite graphs (user–device, user–merchant, and user–address), aggregates the information of the user’s neighbor nodes under the three types of behavioral patterns, and integrates the graph convolution classification results under the three behavioral patterns with the XGBoost model to achieve fraudulent user detection with integrated multiple relational graphs. The results show that the performance of this model in fraud identification is significantly improved, especially in reducing the fraudulent user underreporting rate. Further, the behavioral associations that play a key role in fraud user identification are analyzed in conjunction with shape value to provide a reference for fraud pattern mining.

Spatiotemporal cortical dynamics for visual scene processing as revealed by EEG decoding.

The human visual system rapidly recognizes the categories and global properties of complex natural scenes. The present study investigated the spatiotemporal dynamics of neural signals involved in visual scene processing using electroencephalography (EEG) decoding. We recorded visual evoked potentials from 11 human observers for 232 natural scenes, each of which belonged to one of 13 natural scene categories (e.g., a bedroom or open country) and had three global properties (naturalness, openness, and roughness). We trained a deep convolutional classification model of the natural scene categories and global properties using EEGNet. Having confirmed that the model successfully classified natural scene categories and the three global properties, we applied Grad-CAM to the EEGNet model to visualize the EEG channels and time points that contributed to the classification. The analysis showed that EEG signals in the occipital electrodes at short latencies (approximately 80 ~ ms) contributed to the classifications, whereas those in the frontal electrodes at relatively long latencies (200 ~ ms) contributed to the classification of naturalness and the individual scene category. These results suggest that different global properties are encoded in different cortical areas and with different timings, and that the combination of the EEGNet model and Grad-CAM can be a tool to investigate both temporal and spatial distribution of natural scene processing in the human brain.

A novel wavelet-transform-based convolution classification network for cervical lymph node metastasis of papillary thyroid carcinoma in ultrasound images.

Preoperative assessment of cervical lymph nodes metastasis (CLNM) for accurate qualitative and locating diagnosis is important for choosing the best treatment option for patients with papillary thyroid cancer. Non-destructive, non-invasive ultrasound is currently the imaging method of choice for lymph node metastatic assessment. For lymph node characteristics and ultrasound images, this paper proposes a multitasking network framework for diagnosing metastatic lymph nodes in ultrasound images, in which localization module not only provides information on the location of lymph nodes to focus on the peripheral and self regions of lymph nodes, but also provides structural features of lymph nodes for subsequent classification module. In the classification module, we design a novel wavelet-transform-based convolution network. Wavelet transform is introduced into the deep learning convolution module to analyze ultrasound images in both spatial and frequency domains, which effectively enriches the feature information and improves the classification performance of the model without increasing the model parameters. We collected 510 patient data (N = 1376) from Shanghai Sixth People's Hospital regarding ultrasound lymph nodes in the neck, as well as used three publicly available ultrasound datasets, including SCUI2020 (N = 2914), DDTI (N = 480), and BUSI (N = 780). Compared to the optimal two-stage model, our model has improved its accuracy and AUC indexes by 5.83% and 4%, which outperforms the two-stage architectures and also surpasses the latest classification networks.

3D Point Cloud Classification Method Based on Multiple Attention Mechanism and Dynamic Graph Convolution

In order to solve the problem of uneven density and the low classification accuracy of 3D point cloud, a 3D point cloud classification method fuses multi-attention machine is proposed. It is principally based on the traditional point cloud dynamic graph convolution classification network, into multiple attention mechanisms, including self-attention, spatial attention and channel attention mechanisms. The self-attention mechanism can reduce the dependence on irrelevant points while aligning point clouds, and input the processed point cloud into the classification network. Then the missing geometric information in the classification network is compensated by the integration of spatial and channel attention mechanisms. The experimental results on the public data set ModelNet40 indicate that compared with the DGCNN classification network, the improved network model improves the classification accuracy of the data set by 0.5 % and the average accuracy by 0.9 %. Meantime, the classification accuracy outstrips other contrast classification algorithms.

Comparison of one- two- and three-dimensional CNN models for drawing-test-based diagnostics of the Parkinson’s disease

Subject:In this article, convolutional networks of one, two, and three dimensions are compared with respect to their ability to distinguish between the drawing tests produced by Parkinson’s disease patients and healthy control subjects. Motivation:The application of deep learning techniques for the analysis of drawing tests to support the diagnosis of Parkinson’s disease has become a growing trend in the area of Artificial Intelligence. Methods:The dynamic features of the handwriting signal are embedded in the static test data to generate one-dimensional time series, two-dimensional RGB images and three-dimensional voxelized point clouds, and then one-, two-, and three-dimensional CNN can be used to automatically extract features for effective diagnosis. Novelty:While there are many results that describe the application of two-dimensional convolutional models to the problem, to the best knowledge of the authors, there are no results based on the application of three-dimensional models and very few using one-dimensional models. Main result:The accuracy of the one-, two- and three-dimensional CNN models was 59.38%, 77.73% and 82.34% in the DraWritePD dataset (acquired by the authors) and 63.33%, 81.33% and 82.22% in the PaHaW dataset (well known from the literature), respectively. For these two data sets, the proposed three-dimensional convolutional classification method exhibits the best diagnostic performance.

TasLA: An innovative Tasmanian and Lichtenberg optimized attention deep convolution based data fusion model for IoMT smart healthcare

The Internet of Medical Things (IoMT) bolstered the smart health care industry in present times by enabling quicker patient monitoring and disease diagnosis. However, there have been problems that need to be resolved using Artificial Intelligence (AI) methods. The major goal of this endeavor is to develop an IoMT-based data fusion system for multi-sensor smart healthcare network. To do this, a new optimization and deep learning approaches are being used in this work. In this research work, a unique smart healthcare framework, Tasmanian and Lichtenberg Optimized Attention Deep Convolution (TasLA) is developed for IoMT systems. This system uses an intelligent data fusion algorithms for collecting of medical data and the diagnosis of disorders. Here, data pretreatment and normalization processes are carried out in order to provide a dataset with balanced attribute information. The qualities or characteristics that will aid in classification are then selected using the most modern Tasmanian Devil Optimization (TDO) approach. The Attention Deep Convolution Classification (ADCC) algorithm is also used to classify the medical condition, thereby improving classification precision and reducing false predictions. To optimally compute the loss function during prediction, the Lichtenberg Optimization (LO) technique is employed to enhance classification performance. The effectiveness and results of the proposed TasLA model are validated and contrasted using various benchmark datasets such as Hungarian, Cleveland, Echocardiogram, and Z-Alizadeh.

Weakly Supervised MRI Slice-Level Deep Learning Classification of Prostate Cancer Approximates Full Voxel- and Slice-Level Annotation: Effect of Increasing Training Set Size.

Weakly supervised learning promises reduced annotation effort while maintaining performance. To compare weakly supervised training with full slice-wise annotated training of a deep convolutional classification network (CNN) for prostate cancer (PC). Retrospective. One thousand four hundred eighty-nine consecutive institutional prostate MRI examinations from men with suspicion for PC (65 ± 8 years) between January 2015 and November 2020 were split into training (N = 794, enriched with 204 PROSTATEx examinations) and test set (N = 695). 1.5 and 3T, T2-weighted turbo-spin-echo and diffusion-weighted echo-planar imaging. Histopathological ground truth was provided by targeted and extended systematic biopsy. Reference training was performed using slice-level annotation (SLA) and compared to iterative training utilizing patient-level annotations (PLAs) with supervised feedback of CNN estimates into the next training iteration at three incremental training set sizes (N = 200, 500, 998). Model performance was assessed by comparing specificity at fixed sensitivity of 0.97 [254/262] emulating PI-RADS ≥ 3, and 0.88-0.90 [231-236/262] emulating PI-RADS ≥ 4 decisions. Receiver operating characteristic (ROC) and area under the curve (AUC) was compared using DeLong and Obuchowski test. Sensitivity and specificity were compared using McNemar test. Statistical significance threshold was P = 0.05. Test set (N = 695) ROC-AUC performance of SLA (trained with 200/500/998 exams) was 0.75/0.80/0.83, respectively. PLA achieved lower ROC-AUC of 0.64/0.72/0.78. Both increased performance significantly with increasing training set size. ROC-AUC for SLA at 500 exams was comparable to PLA at 998 exams (P = 0.28). ROC-AUC was significantly different between SLA and PLA at same training set sizes, however the ROC-AUC difference decreased significantly from 200 to 998 training exams. Emulating PI-RADS ≥ 3 decisions, difference between PLA specificity of 0.12 [51/433] and SLA specificity of 0.13 [55/433] became undetectable (P = 1.0) at 998 exams. Emulating PI-RADS ≥ 4 decisions, at 998 exams, SLA specificity of 0.51 [221/433] remained higher than PLA specificity at 0.39 [170/433]. However, PLA specificity at 998 exams became comparable to SLA specificity of 0.37 [159/433] at 200 exams (P = 0.70). Weakly supervised training of a classification CNN using patient-level-only annotation had lower performance compared to training with slice-wise annotations, but improved significantly faster with additional training data. 3 TECHNICAL EFFICACY: Stage 2.

Autism spectrum disorder diagnosis based on deep unrolling-based spatial constraint representation.

Accurate diagnosis of autism spectrum disorder (ASD) is crucial for effective treatment and prognosis. Functional brain networks (FBNs) constructed from functional magnetic resonance imaging (fMRI) have become a popular tool for ASD diagnosis. However, existing model-driven approaches used to construct FBNs lack the ability to capture potential non-linear relationships between data and labels. Moreover, most existing studies treat the FBNs construction and disease classification as separate steps, leading to large inter-subject variability in the estimated FBNs and reducing the statistical power of subsequent group comparison. To address these limitations, we propose a new approach to FBNs construction called the deep unrolling-based spatial constraint representation (DUSCR) model and integrate it with a convolutional classifier to create an end-to-end framework for ASD recognition. Specifically, the model spatial constraint representation (SCR) is solved using a proximal gradient descent algorithm, and we unroll it into deep networks using the deep unrolling algorithm. Classification is then performed using a convolutional prototype learning model. We evaluated the effectiveness of the proposed method on the ABIDE I dataset and observed a significant improvement in model performance and classification accuracy. The resting state fMRI images are preprocessed into time series data and 3D coordinates of each region of interest. The data are fed into the DUSCR model, a model for building functional brain networks using deep learning instead of traditional models, that we propose, and then the outputs are fed into the convolutional classifier with prototype learning to determine whether the patient has ASD disease.