Network Features Research Articles

Automatic prediction of obstructive sleep apnea in patients with temporomandibular disorder based on multidata and machine learning

Obstructive sleep apnea (OSA) is closely associated with the development and chronicity of temporomandibular disorder (TMD). Given the intricate pathophysiology of both OSA and TMD, comprehensive diagnostic approaches are crucial. This study aimed to develop an automatic prediction model utilizing multimodal data to diagnose OSA among TMD patients. We collected a range of multimodal data, including clinical characteristics, portable polysomnography, X-ray, and MRI data, from 55 TMD patients who reported sleep problems. This data was then analyzed using advanced machine learning techniques. Three-dimensional VGG16 and logistic regression models were used to identify significant predictors. Approximately 53% (29 out of 55) of TMD patients had OSA. Performance accuracy was evaluated using logistic regression, multilayer perceptron, and area under the curve (AUC) scores. OSA prediction accuracy in TMD patients was 80.00–91.43%. When MRI data were added to the algorithm, the AUC score increased to 1.00, indicating excellent capability. Only the obstructive apnea index was statistically significant in predicting OSA in TMD patients, with a threshold of 4.25 events/h. The learned features of the convolutional neural network were visualized as a heatmap using a gradient-weighted class activation mapping algorithm, revealing that it focuses on differential anatomical parameters depending on the absence or presence of OSA. In OSA-positive cases, the nasopharynx, oropharynx, uvula, larynx, epiglottis, and brain region were recognized, whereas in OSA-negative cases, the tongue, nose, nasal turbinate, and hyoid bone were recognized. Prediction accuracy and heat map analyses support the plausibility and usefulness of this artificial intelligence-based OSA diagnosis and prediction model in TMD patients, providing a deeper understanding of regions distinguishing between OSA and non-OSA.

EEG-Based Feature Classification Combining 3D-Convolutional Neural Networks with Generative Adversarial Networks for Motor Imagery.

The adoption of convolutional neural networks (CNNs) for decoding electroencephalogram (EEG)-based motor imagery (MI) in brain-computer interfaces has significantly increased recently. The effective extraction of motor imagery features is vital due to the variability among individuals and temporal states. This study introduces a novel network architecture, 3D-convolutional neural network-generative adversarial network (3D-CNN-GAN), for decoding both within-session and cross-session motor imagery. Initially, EEG signals were extracted over various time intervals using a sliding window technique, capturing temporal, frequency, and phase features to construct a temporal-frequency-phase feature (TFPF) three-dimensional feature map. Generative adversarial networks (GANs) were then employed to synthesize artificial data, which, when combined with the original datasets, expanded the data capacity and enhanced functional connectivity. Moreover, GANs proved capable of learning and amplifying the brain connectivity patterns present in the existing data, generating more distinctive brain network features. A compact, two-layer 3D-CNN model was subsequently developed to efficiently decode these TFPF features. Taking into account session and individual differences in EEG data, tests were conducted on both the public GigaDB dataset and the SHU laboratory dataset. On the GigaDB dataset, our 3D-CNN and 3D-CNN-GAN models achieved two-class within-session motor imagery accuracies of 76.49% and 77.03%, respectively, demonstrating the algorithm's effectiveness and the improvement provided by data augmentation. Furthermore, on the SHU dataset, the 3D-CNN and 3D-CNN-GAN models yielded two-class within-session motor imagery accuracies of 67.64% and 71.63%, and cross-session motor imagery accuracies of 58.06% and 63.04%, respectively. The 3D-CNN-GAN algorithm significantly enhances the generalizability of EEG-based motor imagery brain-computer interfaces (BCIs). Additionally, this research offers valuable insights into the potential applications of motor imagery BCIs.

Breaking the silence: leveraging social interaction data to identify high-risk suicide users online using network analysis and machine learning

Suicidal thought and behavior (STB) is highly stigmatized and taboo. Prone to censorship, yet pervasive online, STB risk detection may be improved through development of uniquely insightful digital markers. Focusing on Sanctioned Suicide, an online pro-choice suicide forum, this work derived 17 egocentric network features to capture dynamics of social interaction and engagement within this uniquely uncensored community. Using network data generated from over 3.2 million unique interactions of N = 192 individuals, n = 48 of which were determined to be highest risk users (HRUs), a machine learning classification model was trained, validated, and tested to predict HRU status. Model prediction dynamics were analyzed using introspection techniques to uncover patterns in feature influence and highlight social phenomena. The model achieved a test AUC = 0.73 ([0.61, 0.85], 95% CI), suggesting that network-based socio-behavioral patterns of online interaction can signal for heightened suicide risk. Transitivity, density, and in-degree centrality were among the most important features driving this performance. Moreover, predicted HRUs tended to be targets of social exchanges with lesser frequency and possessed egocentric networks with “small world” network properties. Through the implementation of an underutilized method on an unlikely data source, findings support future incorporation of network-based social interaction features in descriptive, predictive, and preventative STB research.

White matter structure and derived network properties are used to predict the progression from mild cognitive impairment of older adults to Alzheimer’s disease

ObjectiveTo identify white matter fiber injury and network changes that may lead to mild cognitive impairment (MCI) progression, then a joint model was constructed based on neuropsychological scales to predict high-risk individuals for Alzheimer's disease (AD) progression among older adults with MCI.MethodsA total of 173 MCI patients were included from the Alzheimer's Disease Neuroimaging Initiative(ADNI) database and randomly divided into training and testing cohorts. Forty-five progressed to AD during a 4-year follow-up period. Diffusion tensor imaging (DTI) techniques extracted relevant DTI quantitative features for each patient. In addition, brain networks were constructed based on white matter fiber bundles to extract network property features. Ensemble dimensionality reduction was applied to reduce both DTI quantitative features and network features from the training cohort, and machine learning algorithms were added to construct white matter signature. In addition, 52 patients from the National Alzheimer's Coordinating Center (NACC) database were used for external validation of white matter signature. A joint model was subsequently generated by combining with scale scores, and its performance was evaluated using data from the testing cohort.ResultsBased on multivariate logistic regression, clinical dementia rating and Alzheimer’s disease assessment scales (CDRS and ADAS, respectively) were selected as independent predictive factors. A joint model was constructed in combination with the white matter signature. The AUC, sensitivity, and specificity in the training cohort were 0.938, 0.937, and 0.91, respectively, and the AUC, sensitivity, and specificity in the test cohort were 0.905, 0.923, and 0.872, respectively. The Delong test showed a statistically significant difference between the joint model and CDRS or ADAS scores (P < 0.05), yet no significant difference between the joint model and the white matter signature (P = 0.341).ConclusionThe present results demonstrate that a joint model combining neuropsychological scales can be constructed by using machine learning and DTI technology to identify MCI patients who are at high-risk of progressing to AD.

Considering Appropriate Input Features of Neural Network to Calibrate Option Pricing Models

Considering Appropriate Input Features of Neural Network to Calibrate Option Pricing Models

Label-guided low-rank Approximation for functional brain network learning in identifying subcortical vascular cognitive impairment

Background and ObjectiveFunctional brain network (FBN) has played a pivotal role in unraveling the inherent mechanism of cognition/behavior and investigating neuropsychological disorders. Particularly, growing evidence suggests that FBNs have shown great potential in classifying brain disorders such as subcortical vascular cognitive impairment (SVCI). However, learning a high-quality FBN is still challenging since there is no ground truth. Most traditional methods tend to focus on learning FBNs independently of the downstream task. This practice neglects the valuable label information crucial for accurate classification. Besides, the methodology of constructing brain networks in an isolated, individualistic manner sidesteps the common information among individuals. MethodsTo address these issues, we developed a new FBN joint learning strategy named Label-Guided Low-rank Approximation (LGLA). LGLA integrates label information from training subjects and uses unlabeled testing subjects for auxiliary training, enhancing the discriminative power of FBN features tailored to testing subjects. By enforcing class-specific similarities and differences, and applying a low-rank constraint to capture shared information, LGLA presents an optimization framework that can be effectively solved. ResultsExperimental results demonstrate the effectiveness of our proposed FBN learning method for identifying SVCI. The proposed method achieved an accuracy of 75.70% and an area under the receiver operating characteristic curve (AUC) of 0.8233 in the task of identifying patients with amnestic mild cognitive impairment (aMCI). For identifying patients with non-amnestic mild cognitive impairment (naMCI), it achieved an accuracy of 63.64% and an AUC of 0.7857, while the task of identifying patients with no cognitive impairment (NCI), it achieved an accuracy of 63.64% and an AUC of 0.7917. Additionally, we further explored the brain network features and discovered potential biomarkers for personalized diagnosis. ConclusionThe LGLA method enhances FBN construction and classification accuracy for SVCI diagnosis by effectively integrating the label information and shared structure among subjects.

A Comprehensive Survey on Machine Learning Methods for Handover Optimization in 5G Networks

One of the key features of mobile networks in this age of mobile communication is seamless communication. Handover (HO) is a critical component of next-generation (NG) cellular communication networks, which requires careful management since it poses several risks to quality-of-service (QoS), including a decrease in average throughput and service disruptions. Due to the dramatic rise in base stations (BSs) and connections per unit area brought about by new fifth-generation (5G) network enablers, such as Internet of things (IoT), network densification, and mm-wave communications, HO management has become more challenging. The degree of difficulty is increased in light of the strict criteria that were recently published in the specifications of 5G networks. In order to address these issues more successfully and efficiently, this study has explored and examined intelligent HO optimization strategies using machine learning models. Furthermore, the significant goal of this review is to present the state of cellular networks as they are now, as well as to talk about mobility and home office administration in 5G alongside the overall features of 5G networks. This work presents an overview of machine learning methods in handover optimization and of the various data availability for evaluations. In the final section, the challenges and future research directions are also detailed.

SARD: Fake news detection based on CLIP contrastive learning and multimodal semantic alignment

The automatic detection of multimodal fake news can be used to effectively identify potential risks in cyberspace. Most of the existing multimodal fake news detection methods focus on fully exploiting textual and visual features in news content, thus neglecting the full utilization of news social context features that play an important role in improving fake news detection. To this end, we propose a new fake news detection method based on CLIP contrastive learning and multimodal semantic alignment (SARD). SARD leverages cutting-edge multimodal learning techniques, such as CLIP, and robust cross-modal contrastive learning methods to integrate features of news-oriented heterogeneous information networks (N-HIN) with multi-level textual and visual features into a unified framework for the first time. This framework not only achieves cross-modal alignment between deep textual and visual features but also considers cross-modal associations and semantic alignments across different modalities. Furthermore, SARD enhances fake news detection by aligning semantic features between news content and N-HIN features, an aspect largely overlooked by existing methods. We test and evaluate SARD on three real-world datasets. Experimental results demonstrate that SARD significantly outperforms the twelve state-of-the-art competitors in fake news detection, with an average improvement of 2.89% in Mac.F1 score and 2.13% in accuracy compared to the leading baseline models across three datasets.

Video stalling identification for web live streaming under HTTP-FLV

Live broadcasts have become one of the most popular forms of entertainment. Quality of user Experience (QoE) is a vital quantitative criterion for evaluating user satisfaction while watching live broadcasts, and it is positively correlated with the increase in the income of Internet Service Providers (ISPs). Video stalling identification plays a crucial role in the evaluation of QoE. However, encrypted live streaming hides video content, which makes identifying video stalling challenging. Existing studies primarily detect video stalling in a fixed time interval and focus on high-dimensional features. However, the capacity of the client byte buffer is dynamic, resulting in the stalling and non-stalling existing in a certain and fixed stalling time. In addition, the handling time of abundant features causes further latency. We propose Truncation of Dynamic Bytes and non-linear Integrated Modification based on Double Buffers (DB2) to identify video stalling under HTTP-FLV protocol in various network conditions and live types. We pull real-time video to get client buffer parameters and build a dynamic mapping based on the double buffer between network packets and the video playing states. This allows a more objective and precise evaluation of video stalling. We design a new network feature by creating a non-linear relationship between network packets and the client buffer. This is achieved by combining non-linear convergent distribution with basic traffic features. The feature is fed into a lightweight machine learning model to train the classifier, achieving low processing latency and high identification accuracy. The experimental results show that DB2 can achieve 98.91% stalling identification accuracy with 1.256 ms operation time in a mixture of live video types, outperforming state-of-the-art techniques.

Neural quantile optimization for edge–cloud networking

We seek the best traffic allocation scheme for the edge–cloud networking subject to SD-WAN architecture and burstable billing. First, we formulate a family of quantile-based integer programming problems for a fixed network topology with random parameters describing the traffic demands. Then, to overcome the difficulty caused by the discrete feature, we generalize the Gumbel-softmax reparameterization method to induce an unconstrained continuous optimization problem as a regularized continuation of the discrete problem. Finally, we introduce the Gumbel-softmax sampling neural network to solve optimization problems via unsupervised learning. The neural network structure reflects the edge–cloud networking topology and is trained to minimize the expectation of the cost function for unconstrained continuous optimization problems. The trained network works as an efficient traffic allocation scheme sampler, outperforming the random strategy in feasibility and cost value. Besides testing the quality of the output allocation scheme, we examine the generalization property of the network by increasing the time steps and the number of users. We also feed the solution to existing integer optimization solvers as initial conditions and verify the warm-starts can accelerate the short-time iteration process. The framework is general, and the decoupled feature of the random neural networks is adequate for practical implementations.

The neocortical infrastructure for language involves region-specific patterns of laminar gene expression

The language network of the human brain has core components in the inferior frontal cortex and superior/middle temporal cortex, with left-hemisphere dominance in most people. Functional specialization and interconnectivity of these neocortical regions is likely to be reflected in their molecular and cellular profiles. Excitatory connections between cortical regions arise and innervate according to layer-specific patterns. Here, we generated a gene expression dataset from human postmortem cortical tissue samples from core language network regions, using spatial transcriptomics to discriminate gene expression across cortical layers. Integration of these data with existing single-cell expression data identified 56 genes that showed differences in laminar expression profiles between the frontal and temporal language cortex together with upregulation in layer II/III and/or layer V/VI excitatory neurons. Based on data from large-scale genome-wide screening in the population, DNA variants within these 56 genes showed set-level associations with interindividual variation in structural connectivity between the left-hemisphere frontal and temporal language cortex, and with the brain-related disorders dyslexia and schizophrenia which often involve affected language. These findings identify region-specific patterns of laminar gene expression as a feature of the brain's language network.

MADDPG-Based Deployment Algorithm for 5G Network Slicing

One of the core features of 5G networks is the ability to support multiple services on the same infrastructure, with network slicing being a key technology. However, existing network slicing architectures have limitations in efficiently handling slice requests with different requirements, particularly when addressing high-reliability and high-demand services, where many issues remain unresolved. For example, predicting whether actual physical resources can meet network slice request demands and achieving flexible, on-demand resource allocation for different types of slice requests are significant challenges. To address the need for more flexible and efficient service demands, this paper proposes a 5G network slicing deployment algorithm based on the Multi-Agent Deep Deterministic Policy Gradient (MADDPG). Firstly, a new 5G network slicing deployment system framework is established, which measures resources for three typical 5G network slicing scenarios (eMBB, mMTC, uRLLC) and processes different types of slice requests by predicting slice request traffic. Secondly, by adopting the multi-agent approach of MADDPG, the algorithm enhances cooperation between multiple service requests, decentralizes action selection for requests, and schedules resources separately for the three types of slice requests, thereby optimizing resource allocation. Finally, simulation results demonstrate that the proposed algorithm significantly outperforms existing algorithms in terms of resource efficiency and slice request acceptance rate, showcasing the advantages of multi-agent approaches in slice request handling.

Prevalence of and gene regulatory constraints on transcriptional adaptation in single cells

BackgroundCells and tissues have a remarkable ability to adapt to genetic perturbations via a variety of molecular mechanisms. Nonsense-induced transcriptional compensation, a form of transcriptional adaptation, has recently emerged as one such mechanism, in which nonsense mutations in a gene trigger upregulation of related genes, possibly conferring robustness at cellular and organismal levels. However, beyond a handful of developmental contexts and curated sets of genes, no comprehensive genome-wide investigation of this behavior has been undertaken for mammalian cell types and conditions. How the regulatory-level effects of inherently stochastic compensatory gene networks contribute to phenotypic penetrance in single cells remains unclear.ResultsWe analyze existing bulk and single-cell transcriptomic datasets to uncover the prevalence of transcriptional adaptation in mammalian systems across diverse contexts and cell types. We perform regulon gene expression analyses of transcription factor target sets in both bulk and pooled single-cell genetic perturbation datasets. Our results reveal greater robustness in expression of regulons of transcription factors exhibiting transcriptional adaptation compared to those of transcription factors that do not. Stochastic mathematical modeling of minimal compensatory gene networks qualitatively recapitulates several aspects of transcriptional adaptation, including paralog upregulation and robustness to mutation. Combined with machine learning analysis of network features of interest, our framework offers potential explanations for which regulatory steps are most important for transcriptional adaptation.ConclusionsOur integrative approach identifies several putative hits—genes demonstrating possible transcriptional adaptation—to follow-up on experimentally and provides a formal quantitative framework to test and refine models of transcriptional adaptation.

ExpOmics: a comprehensive web platform empowering biologists with robust multi-omics data analysis capabilities.

High-throughput technologies yield a broad spectrum of multi-omics datasets, which offer unparalleled insights into complex biological systems. However, effectively analyzing this diverse array of data presents challenges, considering factors such as species diversity, data types, costs, and limitations of the available tools. Herein, we present ExpOmics, a comprehensive web platform featuring 7 applications and 4 toolkits, with 28 customizable analysis functions spanning various analyses of differential expression, co-expression, Weighted Gene Co-expression Network Analysis (WGCNA), feature selection, and functional enrichment. ExpOmics allows users to upload and explore multi-omics data without organism restrictions, supporting various expression data, including genes, mRNAs, lncRNAs, miRNAs, circRNAs, piRNAs, and proteins and is compatible with diverse gene nomenclatures and expression values. Moreover, ExpOmics enables users to analyze 22427 transcriptomic datasets of 196 cancer subtypes sourced from 63 projects of The Cancer Genome Atlas Program (TCGA) to identify cancer biomarkers. The analysis results from ExpOmics are presented in high-quality graphical formats suitable for publication and are available for free download. A case study using ExpOmics identified two potential oncogenes, SERPINE1 and SLC43A1, that may regulate colorectal cancer through distinct biological processes. In summary, ExpOmics can serves as a robust platform for global researchers to explore multi-omics data, gain biological insights, and formulate testable hypotheses. ExpOmics is available at http://www.biomedical-web.com/expomics.

Global information aware network with global interaction graph attention for infrared small target detection

AbstractDetecting small targets in infrared images is crucial for ground surveillance and air traffic control. However, distinguishing small infrared targets from similar backgrounds is challenging due to their lack of structural and textural characteristics. To address these challenges, this study proposes a novel global information‐aware network with global interaction graph attention (GIGA) for infrared small target detection. The GIGA consists of a global interaction layer (GILayer), graph attention weights (GAW), and a global relational learning (GRL) module. Specifically, the GILayer dynamically learns global inter‐pixel relationships of small target images by enhancing the dependencies between feature dimensions. The GAW component calculates pixel‐by‐pixel similarity across the entire feature map using graph attention mechanisms, while the GRL module retains critical similarity features in the feature extraction network, thereby facilitating small target detection. Additionally, the multi‐scale context fusion module utilises self‐attention and dilation convolution to complement richer feature details at different scales. Experimental results on both natural and synthetic datasets demonstrate the proposed method's superiority over other state‐of‐the‐art conventional and deep learning approaches in infrared small target detection.

Downstream lingering attention transformer network (DsLATNet) for land use land cover classification: A bicolor deep learning framework

Land Use and Land Cover (LULC) classification is the process of locating and classifying regions of the Earth's surface (land cover) based on their physical attributes and human utilization (land use purpose). It is essential for mapping and monitoring changes in ecosystems, facilitating effective resource management and informed decision-making in land-use planning. Convolutional Neural Network (CNN’s) is incapable of capturing long-range dependencies. The effectiveness of transformers relies on extensive training datasets, yet many satellite datasets have comparatively limited sample sizes. Moreover, due to the numerous factors influencing visual prominence, it is challenging to obtain abundant features from a single-color space. To overcome these limitations, we introduce an innovative bicolor architecture Downstream Lingering Attention Transformer Network (DsLATNet). DsLATNet characterizes remote sensing images on two color spaces viz RGB and HSV by processing the information using Residual Network −50(ResNet-50) backbone network.Extracted features of backbone network are further refined to acquire detailed aware features through Downstream Lingering Feature Pyramid Network (DsLFPN).This is further integrated with a DuoTransformer that utilizes attention mechanisms to obtain long-range dependencies. Eventually, the features are aggregated using residual connections that are dilated in-depth multiple times to produce class labels. The efficacy of the algorithm is evaluated through standard classification accuracy metrics. Experimental results on both the Wuhan Dense Labeling Dataset (WHDLD) and the Gaofen Image Dataset (GID) exhibit that the suggested technique exceeds cutting-edge classification methods. WHDLD achieved an Overall Accuracy (OA) of 86.34 %, Average Accuracy (AA) of 75.62 %, and Kappa coefficient (K) of 80.75, while the GID exhibited superior performance with an OA of 86.43 %, AA of 88.13 %, and K of 81.25.

GMRD

The rapid development of social media has allowed people to access information through multiple channels, but social media has also become a breeding ground for rumors. Rumor detection models can effectively assess the credibility of information. However, current research mainly relies on text or combined text and image features, which may not be sufficient to capture complex feature information. Therefore, this paper proposes a rumor detection model based on the graph convolutional network (GCN) and multi-modal features. The proposed model constructs a knowledge graph (KG) and leverages the GCN to extract complex relationships between its nodes. Then, an interactive attention network is adopted to deeply integrate features. Furthermore, ResNet101 is utilized to extract salient features from images, addressing the challenges related to fully utilizing additional feature information and capturing text and image features at a deeper level to some extent. Multiple experiments conducted on datasets from Twitter and Weibo platforms demonstrate the efficacy of the proposed approach.

The ultrametric backbone is the union of all minimum spanning forests

Minimum spanning trees and forests are powerful sparsification techniques that remove cycles from weighted graphs to minimize total edge weight while preserving node reachability, with applications in computer science, network science, and graph theory. Despite their utility and ubiquity, they have several limitations, including that they are only defined for undirected networks, they significantly alter dynamics on networks, and they do not generally preserve important network features such as shortest distances, shortest path distribution, and community structure. In contrast, distance backbones, which are subgraphs formed by all edges that obey a generalized triangle inequality, are well defined in directed and undirected graphs and preserve those and other important network features. The backbone of a graph is defined with respect to a specified path-length operator that aggregates weights along a path to define its length, thereby associating a cost to indirect connections. The backbone is the union of all shortest paths between each pair of nodes according to the specified operator. One such operator, the max function, computes the length of a path as the largest weight of the edges that compose it (a weakest link criterion). It is the only operator that yields an algebraic structure for computing shortest paths that is consistent with De Morgan's laws. Applying this operator yields the ultrametric backbone of a graph in that (semi-triangular) edges whose weights are larger than the length of an indirect path connecting the same nodes (i.e. those that break the generalized triangle inequality based on max as a path-length operator) are removed. We show that the ultrametric backbone is the union of minimum spanning forests in undirected graphs and provides a new generalization of minimum spanning trees to directed graphs that, unlike minimum equivalent graphs and minimum spanning arborescences, preserves all shortest paths and De Morgan's law consistency.

The spatial structure of the tumor immune microenvironment can explain and predict patient response in high-grade serous carcinoma.

Ovarian cancer is the deadliest gynecological malignancy, and therapeutic options and mortality rates over the last three decades have largely not changed. Recent studies indicate that the composition of the tumor immune microenvironment (TIME) influences patient outcomes. To improve spatial understanding of the TIME, we performed multiplexed ion beam imaging on 83 human high-grade serous carcinoma tumor samples, identifying about 160,000 cells across 23 cell types. For 77 of these samples meeting inclusion criteria, we generated composition features based on cell type proportions, spatial features based on the distances between cell types, and spatial network features representing cell interactions and cell clustering patterns, which we linked to traditional clinical and immunohistochemical variables and patient overall survival (OS) and progression-free survival (PFS) outcomes. Among these features, we found several significant univariate correlations, including B-cell contact with M1 macrophages (OS hazard ratio HR=0.696, p=0.011, PFS HR=0.734, p=0.039). We then used high-dimensional random forest models to evaluate out-of-sample predictive performance for OS and PFS outcomes and to derive relative feature importance scores for each feature. The top model for predicting low or high PFS used TIME composition and spatial features and achieved an average AUC (area under the receiver-operating characteristic curve) score of 0.71. The results demonstrate the importance of spatial structure in understanding how the TIME contributes to treatment outcomes. Furthermore, the present study provides a generalizable roadmap for spatial analyses of the TIME in ovarian cancer research.

Surface texture image classification of carbon/phenolic composites in extreme environments using deep learning

AbstractThe classification of ablation images holds significant practical value in thermal protection structures, as it enables the assessment of heat and corrosion resistance of composites. This paper proposes an image‐based deep learning framework to identify the surface texture of carbon/phenolic composites ablative images. First, ablation experiments and collection of surface texture images of carbon/phenolic composites under different thermal environments were conducted in an electric arc wind tunnel. Then, a deep learning model based on a convolutional neural network (CNN) is developed for ablative image classification. The pre‐trained network is ultimately employed as the input for transfer learning. The network's feature extraction layer is trained using the ImageNet dataset, while the global average pooling addresses specific classification tasks. The test results demonstrate that the proposed method effectively classifies the relatively small surface texture dataset, enhances the classification performance of ablative surface texture with an accuracy of up to 97.6%, and exhibits robustness and generalization capabilities.Highlights The paper proposes a new deep learning classification method for ablative images. A model highly sensitive to small and weak features is built. Transfer learning and data enhancement techniques are introduced into classification.