Standard Neural Network Research Articles

Cellular computation and cognition.

Contemporary neural network models often overlook a central biological fact about neural processing: that single neurons are themselves complex, semi-autonomous computing systems. Both the information processing and information storage abilities of actual biological neurons vastly exceed the simple weighted sum of synaptic inputs computed by the "units" in standard neural network models. Neurons are eukaryotic cells that store information not only in synapses, but also in their dendritic structure and connectivity, as well as genetic "marking" in the epigenome of each individual cell. Each neuron computes a complex nonlinear function of its inputs, roughly equivalent in processing capacity to an entire 1990s-era neural network model. Furthermore, individual cells provide the biological interface between gene expression, ongoing neural processing, and stored long-term memory traces. Neurons in all organisms have these properties, which are thus relevant to all of neuroscience and cognitive biology. Single-cell computation may also play a particular role in explaining some unusual features of human cognition. The recognition of the centrality of cellular computation to "natural computation" in brains, and of the constraints it imposes upon brain evolution, thus has important implications for the evolution of cognition, and how we study it.

A Bayesian deep learning framework for reliable fault diagnosis in wind turbine gearboxes under various operating conditions

Vibration-based fault diagnostics combined with deep learning approaches has promising applications in detecting and diagnosing faults in wind turbine gearboxes. Specifically when time series vibration data is transformed to a 2-dimensional cyclic spectral coherence maps, the accuracy of deep neural networks in classifying faults increases. Nevertheless, standard deep learning techniques are vulnerable to inaccurate predictions when tested with new data originating from unseen faults or unusual operating conditions. To address some of these shortcomings in the context of wind turbine gearboxes, this paper investigates fault diagnostics using Bayesian convolutional neural network which provide accurate results with uncertainty bounds reducing wrong overconfident classifications. The performance of Bayesian and standard neural networks is compared using a simulation-based dataset of acceleration signals generated from a multibody dynamic model of a 5 MW wind turbine. The framework proposed in this paper has relevance to fault detection and diagnosis in other rotating machinery applications.

Hidden classification layers: Enhancing linear separability between classes in neural networks layers

Many Deep Learning approaches are based on variations of standard multi-layer feed-forward neural networks. These are also referred to as deep networks. The basic idea is that each hidden neural layer accomplishes a data transformation which is expected to make the data representation “somewhat more linearly separable” than the previous one to obtain a final data representation which is as linearly separable as possible. However, determining the optimal network parameters for these transformations is a crucial challenge. In this study, we propose a Deep Neural Network architecture (Hidden Classification Layer, HCL) which induces an error function involving the output values of all the network layers. The proposed architecture leads toward solutions where the data representations in the hidden layers exhibit a higher degree of linear separability between classes compared to conventional methods. While similar approaches have been discussed in prior literature, this paper presents a new architecture with a novel error function and conducts an extensive experimental analysis. Furthermore, the architecture can be easily integrated into existing frameworks by simply adding densely connected layers and making a straightforward adjustment to the loss function to account for the output of the added layers. The experiments focus on image classification tasks using four well-established datasets, employing as baselines three widely recognized architectures in the literature. The findings reveal that the proposed approach consistently enhances accuracy on the test sets across all considered cases.

Near Real-Time Syndromic Surveillance of Emergency Department Triage Texts Using Natural Language Processing: Case Study in Febrile Convulsion Detection.

Collecting information on adverse events following immunization from as many sources as possible is critical for promptly identifying potential safety concerns and taking appropriate actions. Febrile convulsions are recognized as an important potential reaction to vaccination in children aged <6 years. The primary aim of this study was to evaluate the performance of natural language processing techniques and machine learning (ML) models for the rapid detection of febrile convulsion presentations in emergency departments (EDs), especially with respect to the minimum training data requirements to obtain optimum model performance. In addition, we examined the deployment requirements for a ML model to perform real-time monitoring of ED triage notes. We developed a pattern matching approach as a baseline and evaluated ML models for the classification of febrile convulsions in ED triage notes to determine both their training requirements and their effectiveness in detecting febrile convulsions. We measured their performance during training and then compared the deployed models' result on new incoming ED data. Although the best standard neural networks had acceptable performance and were low-resource models, transformer-based models outperformed them substantially, justifying their ongoing deployment. Using natural language processing, particularly with the use of large language models, offers significant advantages in syndromic surveillance. Large language models make highly effective classifiers, and their text generation capacity can be used to enhance the quality and diversity of training data.

E-commerce consumption data prediction and privacy protection based on improved neural network algorithm

In today's shopping platform, some enterprises can use all commercial means to obtain users' personal information. How to protect the privacy of consumers has become an important issue. This paper discusses the shortcomings of the traditional neural network algorithm, and puts forward some countermeasures to improve it. Based on the basic principle of standard neural network, a new improved algorithm is developed according to the needs of this paper. The basic principle of the algorithm is based on the added momentum method. This paper mainly uses data models to collect and organize user data through user behavior patterns, so as to prove that the record of target user data cannot predict the possibility of user transportation and consumption, so as to better promote the improvement of transactions and services through these data models.

Parallel learning by multitasking neural networks

Parallel learning, namely the simultaneous learning of multiple patterns, constitutes a modern challenge for neural networks. While this cannot be accomplished by standard Hebbian associative neural networks, in this paper we show how the multitasking Hebbian network (a variation on the theme of the Hopfield model, working on sparse datasets) is naturally able to perform this complex task. We focus on systems processing in parallel a finite (up to logarithmic growth in the size of the network) number of patterns, mirroring the low-storage setting of standard associative neural networks. When patterns to be reconstructed are mildly diluted, the network handles them hierarchically, distributing the amplitudes of their signals as power laws w.r.t. the pattern information content (hierarchical regime), while, for strong dilution, the signals pertaining to all the patterns are simultaneously raised with the same strength (parallel regime). Further, we prove that the training protocol (either supervised or unsupervised) neither alters the multitasking performances nor changes the thresholds for learning. We also highlight (analytically and by Monte Carlo simulations) that a standard cost function (i.e. the Hamiltonian) used in statistical mechanics exhibits the same minima as a standard loss function (i.e. the sum of squared errors) used in machine learning.

The percolating cluster is invisible to image recognition with deep learning

We study the two-dimensional site-percolation model on a square lattice. In this paradigmatic model, sites are randomly occupied with probability p; a second-order phase transition from a non-percolating to a fully percolating phase appears at occupation density pc , called percolation threshold. Through supervised deep learning approaches like classification and regression, we show that standard convolutional neural networks (CNNs), known to work well in similar image recognition tasks, can identify pc and indeed classify the states of a percolation lattice according to their p content or predict their p value via regression. When using instead of p the spatial cluster correlation length ξ as labels, the recognition is beginning to falter. Finally, we show that the same network struggles to detect the presence of a spanning cluster. Rather, predictive power seems lost and the absence or presence of a global spanning cluster is not noticed by a CNN with a local convolutional kernel. Since the existence of such a spanning cluster is at the heart of the percolation problem, our results suggest that CNNs require careful application when used in physics, particularly when encountering less-explored situations.

Computer aided model for lung cancer classification using cat optimized convolutional neural networks

Lung cancer is a dangerous illness in the modern world. Lung cancer diagnosis requires a more thorough investigation than other disease processes since it affects both men and women and has a higher fatality rate. The law currently mandates the creation of sophisticated expert systems for clinical analysis and the precise identification of lung disease therapy. Images from a Computer Tomography (CT) scan can give more useful information about a lung cancer diagnosis. CT scan input pictures are used in the formulation of several Machine Learning (ML) and Deep Learning (DL) algorithms for the enhancement of diagnosis and treatment procedures. The most challenging aspect of research, however, is still creating an accurate and sophisticated system. This study proposes Deep Fused Features-Based Cat-Optimized Networks (DFF-CON), a novel classification approach that utilizes the concepts of fused features and optimized networks. In the proposed structure, Deep Convolutional Neural Networks (DCNN) are employed to enhance the classification maps and ultimately decrease the probability of an overfitting issues. In the framework, saliency maps are used as a first-tier segmentation technique. In order to get the best diagnosis for malignancies in lung CT scan pictures, the suggested study additionally substituted a cat-optimized optimized CNN for the standard neural network. The suggested solution has been developed using Python 3.8, the Keras API, and Tensorflow 1.8 together. With the LIDC-IDRI image datasets, a wide range of performance measures are created and examined for thorough testing, including accuracy, sensitivity, specificity, precision, and the f1-score. Results from the simulation indicate that the suggested framework achieves high accuracy (Ac) (99.89 %), sensitivity (Sy) (99.8 %), specificity (Sy) (99.76 %), precision (Pn) (99.8 %) and F1-score (Fr) (99.88 %). Then, an assessment with other current models is conducted to demonstrate the benefits of the presented framework.

Physics based data driven method for the crashworthiness design of origami composite tubes

A novel method based on a physics informed data driven model is developed to design an origami composite crash tube. The structure consists of two axially stacked basic components, called modules. Each module presents lower and upper square sections with an octagonal section in the middle. The parameters of the octagonal cross-section and the height of each module are optimized to maximize the energy absorption of the tube when subjected to an axial impact. In contrast to standard surrogate modelling techniques, whose accuracy only depends on the amount of available data, a Physics-informed Neural Network (PINN) scheme is adopted to correlate the crushing response of the single modules to that of the whole origami tube, constraining the data driven method to physically consistent predictions. The PINN is first trained on the results obtained with an experimentally validated Finite Element model and then used to optimize the structure. Results show that the PINN can accurately predict the crushing response of the origami tube, while consistently reducing the computational effort required to explore the whole design domain. Also, the comparison with a standard Feed Forward Neural Network (FFNN) shows that the PINN scheme leads to more accurate results.

Cultivating Insight: Detecting Autism Spectrum Disorder through Residual Attention Network in Facial Image Analysis

Revolutionizing Autism Spectrum Disorder Identification through Deep Learning: Unveiling Facial Activation Patterns. In this study, our primary objective is to harness the power of deep learning algorithms for the precise identification of individuals with autism spectrum disorder (ASD) solely from facial image datasets. Our investigation centers around the utilization of face activation patterns, aiming to uncover novel insights into the distinctive facial features of ASD patients. To accomplish this, we meticulously examined facial imaging data from a global and multidisciplinary repository known as the Autism Face Imaging Data Exchange. Autism spectrum disorder is characterized by inherent social deficits and manifests in a spectrum of diverse symptomatic scenarios. Recent data from the Centers for Disease Control (CDC) underscores the significance of this disorder, indicating that approximately 1 in 54 children are impacted by ASD, according to estimations from the CDC's Autism and Developmental Disabilities Monitoring Network (ADDM). Our research delved into the intricate functional connectivity patterns that objectively distinguish ASD participants, focusing on their facial imaging data. Through this investigation, we aimed to uncover the latent facial patterns that play a pivotal role in the classification of ASD cases. Our approach introduces a novel module that enhances the discriminative potential of standard convolutional neural networks (CNNs), such as ResNet-50, thus significantly advancing the state-of-the-art. Our model achieved an impressive accuracy rate of 99% in distinguishing between ASD patients and control subjects within the dataset. Our findings illuminate the specific facial expression domains that contribute most significantly to the differentiation of ASD cases from typically developing individuals, as inferred from our deep learning methodology. To validate our approach, we conducted real-time video testing on diverse children, achieving an outstanding accuracy score of 99.90% and an F1 score of 99.67%. Through this pioneering work, we not only offer a cutting-edge approach to ASD identification but also contribute to the understanding of the underlying facial activation patterns that hold potential for transforming the diagnostic landscape of autism spectrum disorder.

RETRACTED: A Global Structural Hypergraph Convolutional Model for Bundle Recommendation

Bundle recommendations provide personalized suggestions to users by combining related items into bundles, aiming to enhance users’ shopping experiences and boost merchants’ sales revenue. Existing solutions based on graph neural networks (GNN) face several significant challenges: (1) it is demanding to explicitly model multiple complex associations using standard graph neural networks, (2) numerous additional nodes and edges are introduced to approximate higher-order associations, and (3) the user–bundle historical interaction data are highly sparse. In this work, we propose a global structural hypergraph convolutional model for bundle recommendation (SHCBR) to address the above problems. Specifically, we jointly incorporate multiple complex interactions between users, items, and bundles into a relational hypergraph without introducing additional nodes and edges. The hypergraph structure inherently incorporates higher-order associations, thereby alleviating the training burden on neural networks and the dilemma of scarce data effectively. In addition, we design a special matrix propagation rule that captures non-pairwise complex relationships between entities. Using item nodes as links, structural hypergraph convolutional networks learn representations of users and bundles on a relational hypergraph. Experiments conducted on two real-world datasets demonstrate that the SHCBR outperforms the state-of-the-art baselines by 11.07–25.66% on Recall and 16.81–33.53% on NDCG. Experimental results further indicate that the approach based on hypergraphs can offer new insights for addressing bundle recommendation challenges. The codes and datasets have been publicly released on GitHub.

Tac2Pose: Tactile object pose estimation from the first touch

In this paper, we present Tac2Pose, an object-specific approach to tactile pose estimation from the first touch for known objects. Given the object geometry, we learn a tailored perception model in simulation that estimates a probability distribution over possible object poses given a tactile observation. To do so, we simulate the contact shapes that a dense set of object poses would produce on the sensor. Then, given a new contact shape obtained from the sensor, we match it against the pre-computed set using an object-specific embedding learned using contrastive learning. We obtain contact shapes from the sensor with an object-agnostic calibration step that maps RGB (red, green, blue) tactile observations to binary contact shapes. This mapping, which can be reused across object and sensor instances, is the only step trained with real sensor data. This results in a perception model that localizes objects from the first real tactile observation. Importantly, it produces pose distributions and can incorporate additional pose constraints coming from other perception systems, multiple contacts, or priors. We provide quantitative results for 20 objects. Tac2Pose provides high accuracy pose estimations from distinctive tactile observations while regressing meaningful pose distributions to account for those contact shapes that could result from different object poses. We extend and test Tac2Pose in multi-contact scenarios where two tactile sensors are simultaneously in contact with the object, as during a grasp with a parallel jaw gripper. We further show that when the output pose distribution is filtered with a prior on the object pose, Tac2Pose is often able to improve significantly on the prior. This suggests synergistic use of Tac2Pose with additional sensing modalities (e.g., vision) even in cases where the tactile observation from a grasp is not sufficiently discriminative. Given a coarse estimate of an object’s pose, even ambiguous contacts can be used to determine an object’s pose precisely. We also test Tac2Pose on object models reconstructed from a 3D scanner, to evaluate the robustness to uncertainty in the object model. We show that even in the presence of model uncertainty, Tac2Pose is able to achieve fine accuracy comparable to when the object model is the manufacturer’s CAD (computer-aided design) model. Finally, we demonstrate the advantages of Tac2Pose compared with three baseline methods for tactile pose estimation: directly regressing the object pose with a neural network, matching an observed contact to a set of possible contacts using a standard classification neural network, and direct pixel comparison of an observed contact with a set of possible contacts. Website: mcube.mit.edu/research/tac2pose.html

Lightning risk assessment of offshore wind farms by semi-supervised learning

The wind turbine has rapidly developed worldwide with increasing height and scale, resulting in the increased risk of lightning strikes. When wind turbines were stroke by the lightning, they will be damaged, causing economic loss and outage. Lightning risk assessment can guide the improvement of lightning protection and the design of the wind farms to efficiently prevent lightning damages. The traditional lightning risk assessment methods rely on subjective features to some extent. The existing lightning risk assessment methods based on machine learning demand abundant labeled data. It is extremely difficult to label and acquire the data. This paper proposed a lightning risk assessment method based on semi-supervised learning to address the challenges of labeling negative samples and limited labeled data. The semi-supervised K-means algorithm is proposed to divide all data into three parts. The Laplacian support vector machine (LapSVM) with hyperparameters optimized by the particle swarm optimization (PSO) is used to assess the lightning risk. The proposed method has better performance than the standard SVM and neural network (NN). Moreover, previous researches did not consider lightning protection ability. This paper introduces the receptor number into lightning risk assessment. The assessment results suggest that there is higher lightning risk in areas with a great number of wind turbines, high lightning density, and strong lightning strength. The ocean area is more likely to have low lightning risk. The results are valuable for lightning protection optimization of existing wind farms and can give guidance for the plan of new wind farms.

An explainable deep-learning model to stage sleep states in children and propose novel EEG-related patterns in sleep apnea

Automatic deep-learning models used for sleep scoring in children with obstructive sleep apnea (OSA) are perceived as black boxes, limiting their implementation in clinical settings. Accordingly, we aimed to develop an accurate and interpretable deep-learning model for sleep staging in children using single-channel electroencephalogram (EEG) recordings. We used EEG signals from the Childhood Adenotonsillectomy Trial (CHAT) dataset (n = 1637) and a clinical sleep database (n = 980). Three distinct deep-learning architectures were explored to automatically classify sleep stages from a single-channel EEG data. Gradient-weighted Class Activation Mapping (Grad-CAM), an explainable artificial intelligence (XAI) algorithm, was then applied to provide an interpretation of the singular EEG patterns contributing to each predicted sleep stage. Among the tested architectures, a standard convolutional neural network (CNN) demonstrated the highest performance for automated sleep stage detection in the CHAT test set (accuracy = 86.9% and five-class kappa = 0.827). Furthermore, the CNN-based estimation of total sleep time exhibited strong agreement in the clinical dataset (intra-class correlation coefficient = 0.772). Our XAI approach using Grad-CAM effectively highlighted the EEG features associated with each sleep stage, emphasizing their influence on the CNN's decision-making process in both datasets. Grad-CAM heatmaps also allowed to identify and analyze epochs within a recording with a highly likelihood to be misclassified, revealing mixed features from different sleep stages within these epochs. Finally, Grad-CAM heatmaps unveiled novel features contributing to sleep scoring using a single EEG channel. Consequently, integrating an explainable CNN-based deep-learning model in the clinical environment could enable automatic sleep staging in pediatric sleep apnea tests.

Physics informed neural network for charged particles surrounded by conductive boundaries

Molecular dynamics of charged particles in porous conductive media have received considerable attention in recent years due to their application in cutting-edge technologies such as batteries and supercapacitors. Due to the presence of long-range electrical interactions, induced charges present at the boundary, and the influence of boundary conditions, the simulation of these systems is more challenging than the simulation of typical molecular dynamic systems. Simulating these kinds of systems typically involves using a numerical solver to solve the Poisson equation, which is a very time-consuming procedure. Recently, Physics-Informed Neural Networks (PINNs) have been introduced as an alternative to numerical solutions of PDEs. In this paper, we present a new PINN-based model for predicting the potential of point-charged particles surrounded by conductive walls. As a result of the proposed PINN model, the mean square error is less than 7% and R^{2} score is more than 90% for the corresponding example simulation. Results have been compared with typical neural networks and random forest as standard machine learning algorithms. The R^{2} score of the random forest model was 70%, and a standard neural network could not be trained well.

Deep ensemble learning for high-dimensional subsurface fluid flow modeling

The accuracy of Deep Learning (DL) algorithms can be improved by combining several deep learners into an ensemble. This avoids the continuous endeavor required to adjust the architecture of individual networks or the nature of the propagation. This study investigates prediction improvements possible using Deep Ensemble Learning (DEL) to determine four distinct multiscale basis functions in the mixed Generalized Multiscale Finite Element Method (GMsFEM), involving the permeability field as the only input. 376,250 samples were initially generated, filtered down to 367,811 after data pre-processing. A standard Convolutional Neural Network (CNN) named SkiplessCNN and three skip connection-based CNNs named FirstSkipCNN, MidSkipCNN, and DualSkipCNN were developed for the base learners. For each basis function, these four CNNs were combined into an ensemble model using linear regression and ridge regression, separately, as part of the stacking technique. A comparison of the coefficient of determination (R2) and Mean Squared Error (MSE) confirms the effectiveness of all three skip connections in enhancing the performance of the standard CNN, with DualSkip being the most effective among them. Additionally, as evaluated on the testing subset, the combined models meaningfully outperform the individual models for all basis functions. The case that applies linear regression delivers R2 ranging from 0.8456 to 0.9191 and MSE ranging from 0.0092 to 0.0369. The ridge regression case achieves marginally better predictions with R2 ranging from 0.8539 to 0.922, and MSE ranging from 0.009 to 0.0349 because its solution involves more evenly distributed weights.

The Predictive Power of Social Media Sentiment: Evidence from Cryptocurrencies and Stock Markets Using NLP and Stochastic ANNs

Cryptocurrencies are nowadays seen as an investment opportunity, since they show some peculiar features, such as high volatility and diversification properties, that are triggering research interest into investigating their differences with traditional assets. In our paper, we address the problem of predictability of cryptocurrency and stock trends by using data from social online communities and platforms to assess their contribution in terms of predictive power. We extend recent developments in the field by exploiting a combination of stochastic neural networks (NNs), an extension of standard NNs, natural language processing (NLP) to extract sentiment from Twitter, and an external evolutionary algorithm for optimal parameter setting to predict the short-term trend direction. Our results point to good and robust accuracy over time and across different market regimes. Furthermore, we propose to exploit recent advances in sentiment analysis to reassess its role in financial forecasting; in this way, we contribute to the empirical literature by showing that predictions based on sentiment analysis are not found to be significantly different from predictions based on historical data. Nonetheless, compared to stock markets, we find that the accuracy of trend predictions with sentiment analysis is on average much higher for cryptocurrencies.

Fusion of Centroid-Based Clustering With Graph Clustering: An Expectation-Maximization-Based Hybrid Clustering.

This article extends the expectation-maximization (EM) formulation for the Gaussian mixture model (GMM) with a novel weighted dissimilarity loss. This extension results in the fusion of two different clustering methods, namely, centroid-based clustering and graph clustering in the same framework in order to leverage their advantages. The fusion of centroid-based clustering and graph clustering results in a simple "soft" asynchronous hybrid clustering method. The proposed algorithm may start as a pure centroid-based clustering algorithm (e.g., k -means), and as the time evolves, it may eventually and gradually turn into a pure graph clustering algorithm [e.g., basic greedy asynchronous distributed interference avoidance (GADIA) (Babadi and Tarokh, 2010)] as the algorithm converges and vice versa. The "hard" version of the proposed hybrid algorithm includes the standard Hopfield neural networks (and, thus, Bruck's Ln algorithm by (Bruck, 1990) and the Ising model in statistical mechanics), Babadi and Tarokh's basic GADIA in 2010, and the standard k -means (Steinhaus, 1956), (MacQueen, 1967) [i.e., the Lloyd algorithm (Lloyd, 1957, 1982)] as its special cases. We call the "hard version" of the proposed clustering as "hybrid-nongreedy asynchronous clustering (H-NAC)." We apply the H-NAC to various clustering problems using well-known benchmark datasets. The computer simulations confirm the superior performance of the H-NAC compared to the k -means clustering, k -GADIA, spectral clustering, and a very recent clustering algorithm structured graph learning (SGL) by Kang et al. (2021), which represents one of the state-of-the-art clustering algorithms.

A Novel Classification Model Using Optimal Long Short-Term Memory for Classification of COVID-19 from CT Images.

The human respiratory system is affected when an individual is infected with COVID-19, which became a global pandemic in 2020 and affected millions of people worldwide. However, accurate diagnosis of COVID-19 can be challenging due to small variations in typical and COVID-19 pneumonia, as well as the complexities involved in classifying infection regions. Currently, various deep learning (DL)-based methods are being introduced for the automatic detection of COVID-19 using computerized tomography (CT) scan images. In this paper, we propose the pelican optimization algorithm-based long short-term memory (POA-LSTM) method for classifying coronavirus using CT scan images. The data preprocessing technique is used to convert raw image data into a suitable format for subsequent steps. Here, we develop a general framework called no new U-Net (nnU-Net) for region of interest (ROI) segmentation in medical images. We apply a set of heuristic guidelines derived from the domain to systematically optimize the ROI segmentation task, which represents the dataset's key properties. Furthermore, high-resolution net (HRNet) is a standard neural network design developed for feature extraction. HRNet chooses the top-down strategy over the bottom-up method after considering the two options. It first detects the subject, generates a bounding box around the object and then estimates the relevant feature. The POA is used to minimize the subjective influence of manually selected parameters and enhance the LSTM's parameters. Thus, the POA-LSTM is used for the classification process, achieving higher performance for each performance metric such as accuracy, sensitivity, F1-score, precision, and specificity of 99%, 98.67%, 98.88%, 98.72%, and 98.43%, respectively.

A Space-Refine Paradigm for Automatic Carotid Artery Centerline Extraction in Magnetic Resonance Imaging.

Vessel centerline extraction is essential for carotid stenosis assessment and atherosclerotic plaque identification in clinical diagnosis. Simultaneously, it provides a region of interest identification and boundary initialization for computer-assisted diagnosis tools. In magnetic resonance imaging (MRI) cross-sectional images, the lumen shape and vascular topology result in a challenging task to extract the centerline accurately. To this end, we propose a space-refine framework, which exploits the positional continuity of the carotid artery from frame to frame to extract the carotid artery centerline. The proposed framework consists of a detector and a refinement module. Specifically, the detector roughly extracts the carotid lumen region from the original image. Then, we introduce a refinement module that uses the cascade of regressors from a detector to perform sequence realignment of lumen bounding boxes for each subject. It improves the lumen localization results and further enhances the centerline extraction accuracy. Verified by large carotid artery data, the proposed framework achieves state-of-the-art performance compared to conventional vessel centerline extraction methods or standard convolutional neural network approaches.Clinical relevance- Our proposed framework can be used as an important aid for physicians to quantitatively analyze the carotid artery in clinical practice. It is also used as a new paradigm for extracting the centerline of carotid vessels in computer-assisted tools.