Early identification of stroke through deep learning with multi-modal human speech and movement data

Early identification and treatment of stroke can greatly improve patient outcomes and quality of life.Although clinical tests such as the Cincinnati Pre-hospital Stroke Scale(CPSS)and the Face Arm Speech Test(FAST)are commonly used for stroke screening,accurate administration is dependent on specialized training.In this study,we proposed a novel multimodal deep learning approach,based on the FAST,for assessing suspected stroke patients exhibiting symptoms such as limb weakness,facial paresis,and speech disorders in acute settings.We collected a dataset comprising videos and audio recordings of emergency room patients performing designated limb movements,facial expressions,and speech tests based on the FAST.We compared the constructed deep learning model,which was designed to process multi-modal datasets,with six prior models that achieved good action classification performance,including the I3D,SlowFast,X3D,TPN,TimeSformer,and MViT.We found that the findings of our deep learning model had a higher clinical value compared with the other approaches.Moreover,the multi-modal model outperformed its single-module variants,highlighting the benefit of utilizing multiple types of patient data,such as action videos and speech audio.These results indicate that a multi-modal deep learning model combined with the FAST could greatly improve the accuracy and sensitivity of early stroke identification of stroke,thus providing a practical and powerful tool for assessing stroke patients in an emergency clinical setting.

Advancements in Liver Tumor Detection:A Comprehensive Review of Various Deep Learning Models

Liver cancer remains a leading cause of mortality worldwide,and precise diagnostic tools are essential for effective treatment planning.Liver Tumors(LTs)vary significantly in size,shape,and location,and can present with tissues of similar intensities,making automatically segmenting and classifying LTs from abdominal tomography images crucial and challenging.This review examines recent advancements in Liver Segmentation(LS)and Tumor Segmentation(TS)algorithms,highlighting their strengths and limitations regarding precision,automation,and resilience.Performance metrics are utilized to assess key detection algorithms and analytical methods,emphasizing their effectiveness and relevance in clinical contexts.The review also addresses ongoing challenges in liver tumor segmentation and identification,such as managing high variability in patient data and ensuring robustness across different imaging conditions.It suggests directions for future research,with insights into technological advancements that can enhance surgical planning and diagnostic accuracy by comparing popular methods.This paper contributes to a comprehensive understanding of current liver tumor detection techniques,provides a roadmap for future innovations,and improves diagnostic and therapeutic outcomes for liver cancer by integrating recent progress with remaining challenges.

An Enhanced Lung Cancer Detection Approach Using Dual-Model Deep Learning Technique

Lung cancer continues to be a leading cause of cancer-related deaths worldwide,emphasizing the critical need for improved diagnostic techniques.Early detection of lung tumors significantly increases the chances of successful treatment and survival.However,current diagnostic methods often fail to detect tumors at an early stage or to accurately pinpoint their location within the lung tissue.Single-model deep learning technologies for lung cancer detection,while beneficial,cannot capture the full range of features present in medical imaging data,leading to incomplete or inaccurate detection.Furthermore,it may not be robust enough to handle the wide variability in medical images due to different imaging conditions,patient anatomy,and tumor characteristics.To overcome these disadvantages,dual-model or multi-model approaches can be employed.This research focuses on enhancing the detection of lung cancer by utilizing a combination of two learning models:a Convolutional Neural Network(CNN)for categorization and the You Only Look Once(YOLOv8)architecture for real-time identification and pinpointing of tumors.CNNs automatically learn to extract hierarchical features from raw image data,capturing patterns such as edges,textures,and complex structures that are crucial for identifying lung cancer.YOLOv8 incorporates multiscale feature extraction,enabling the detection of tumors of varying sizes and scales within a single image.This is particularly beneficial for identifying small or irregularly shaped tumors that may be challenging to detect.Furthermore,through the utilization of cutting-edge data augmentation methods,such as Deep Convolutional Generative Adversarial Networks(DCGAN),the suggested approach can handle the issue of limited data and boost the models’ability to learn from diverse and comprehensive datasets.The combined method not only improved accuracy and localization but also ensured efficient real-time processing,which is crucial for practical clinical applications.The CNN achieved an accuracy of 97.67%in classifying lung tissues into healthy and cancerous categories.The YOLOv8 model achieved an Intersection over Union(IoU)score of 0.85 for tumor localization,reflecting high precision in detecting and marking tumor boundaries within the images.Finally,the incorporation of synthetic images generated by DCGAN led to a 10%improvement in both the CNN classification accuracy and YOLOv8 detection performance.

Comparative analysis of empirical and deep learning models for ionospheric sporadic E layer prediction

Sporadic E(Es)layers in the ionosphere are characterized by intense plasma irregularities in the E region at altitudes of 90-130 km.Because they can significantly influence radio communications and navigation systems,accurate forecasting of Es layers is crucial for ensuring the precision and dependability of navigation satellite systems.In this study,we present Es predictions made by an empirical model and by a deep learning model,and analyze their differences comprehensively by comparing the model predictions to satellite RO measurements and ground-based ionosonde observations.The deep learning model exhibited significantly better performance,as indicated by its high coefficient of correlation(r=0.87)with RO observations and predictions,than did the empirical model(r=0.53).This study highlights the importance of integrating artificial intelligence technology into ionosphere modelling generally,and into predicting Es layer occurrences and characteristics,in particular.

Machine learning and deep learning to improve prevention of anastomotic leak after rectal cancer surgery

Anastomotic leakage(AL)is a significant complication following rectal cancer surgery,adversely affecting both quality of life and oncological outcomes.Recent advancements in artificial intelligence(AI),particularly machine learning and deep learning,offer promising avenues for predicting and preventing AL.These technologies can analyze extensive clinical datasets to identify preoperative and perioperative risk factors such as malnutrition,body composition,and radiological features.AI-based models have demonstrated superior predictive power compared to traditional statistical methods,potentially guiding clinical decisionmaking and improving patient outcomes.Additionally,AI can provide surgeons with intraoperative feedback on blood supply and anatomical dissection planes,minimizing the risk of intraoperative complications and reducing the likelihood of AL development.

Deep Learning and Artificial Intelligence-Driven Advanced Methods for Acute Lymphoblastic Leukemia Identification and Classification: A Systematic Review

Automatic detection of Leukemia or blood cancer is one of the most challenging tasks that need to be addressed in the healthcare system.Analysis of white blood cells(WBCs)in the blood or bone marrow microscopic slide images play a crucial part in early identification to facilitate medical experts.For Acute Lymphocytic Leukemia(ALL),the most preferred part of the blood or marrow is to be analyzed by the experts before it spreads in the whole body and the condition becomes worse.The researchers have done a lot of work in this field,to demonstrate a comprehensive analysis few literature reviews have been published focusing on various artificial intelligence-based techniques like machine and deep learning detection of ALL.The systematic review has been done in this article under the PRISMA guidelines which presents the most recent advancements in this field.Different image segmentation techniques were broadly studied and categorized from various online databases like Google Scholar,Science Direct,and PubMed as image processing-based,traditional machine and deep learning-based,and advanced deep learning-based models were presented.Convolutional Neural Networks(CNN)based on traditional models and then the recent advancements in CNN used for the classification of ALL into its subtypes.A critical analysis of the existing methods is provided to offer clarity on the current state of the field.Finally,the paper concludes with insights and suggestions for future research,aiming to guide new researchers in the development of advanced automated systems for detecting life-threatening diseases.

Physically Constrained Adaptive Deep Learning for Ocean Vertical-Mixing Parameterization

Existing traditional ocean vertical-mixing schemes are empirically developed without a thorough understanding of the physical processes involved,resulting in a discrepancy between the parameterization and forecast results.The uncertainty in ocean-mixing parameterization is primarily responsible for the bias in ocean models.Benefiting from deep-learning technology,we design the Adaptive Fully Connected Module with an Inception module as the baseline to minimize bias.It adaptively extracts the best features through fully connected layers with different widths,and better learns the nonlinear relationship between input variables and parameterization fields.Moreover,to obtain more accurate results,we impose KPP(K-Profile Parameterization)and PP(Pacanowski–Philander)schemes as physical constraints to make the network parameterization process follow the basic physical laws more closely.Since model data are calculated with human experience,lacking some unknown physical processes,which may differ from the actual data,we use a decade-long time record of hydrological and turbulence observations in the tropical Pacific Ocean as training data.Combining physical constraints and a nonlinear activation function,our method catches its nonlinear change and better adapts to the oceanmixing parameterization process.The use of physical constraints can improve the final results.

A Deep Learning Estimation Method for Temperature-Induced Girder End Displacements of Suspension Bridges

To improve the accuracy of thermal response estimation and overcome the limitations of the linear regression model and Artificial Neural Network(ANN)model,this study introduces a deep learning estimation method specifically based on the Long Short-Term Memory(LSTM)network,to predict temperature-induced girder end displacements of the Dasha Waterway Bridge,a suspension bridge in China.First,to enhance data quality and select target sensors,preprocessing based on the sigma rule and nearest neighbor interpolation is applied to the raw data.Furthermore,to eliminate the high-frequency components from the displacement signal,the wavelet transform is conducted.Subsequently,the linear regression model and ANN model are established,whose results do not meet the requirements and fail to address the time lag effect between temperature and displacements.The study proceeds to develop the LSTM network model and determine the optimal parameters through hyperparameter sensitivity analysis.Finally,the results of the LSTM network model are discussed by a comparative analysis against the linear regression model and ANN model,which indicates a higher accuracy in predicting temperatureinduced girder end displacements and the ability to mitigate the time-lag effect.To be more specific,in comparison between the linear regression model and LSTM network,the mean square error decreases from 6.5937 to 1.6808 and R2 increases from 0.683 to 0.930,which corresponds to a 74.51%decrease in MSE and a 36.14%improvement in R2.Compared to ANN,with an MSE of 4.6371 and an R2 of 0.807,LSTM shows a decrease in MSE of 63.75%and an increase in R2 of 13.23%,demonstrating a significant enhancement in predictive performance.

How Do Deep Learning Forecasting Models Perform for Surface Variables in the South China Sea Compared to Operational Oceanography Forecasting Systems?

It is fundamental and useful to investigate how deep learning forecasting models(DLMs)perform compared to operational oceanography forecast systems(OFSs).However,few studies have intercompared their performances using an identical reference.In this study,three physically reasonable DLMs are implemented for the forecasting of the sea surface temperature(SST),sea level anomaly(SLA),and sea surface velocity in the South China Sea.The DLMs are validated against both the testing dataset and the“OceanPredict”Class 4 dataset.Results show that the DLMs'RMSEs against the latter increase by 44%,245%,302%,and 109%for SST,SLA,current speed,and direction,respectively,compared to those against the former.Therefore,different references have significant influences on the validation,and it is necessary to use an identical and independent reference to intercompare the DLMs and OFSs.Against the Class 4 dataset,the DLMs present significantly better performance for SLA than the OFSs,and slightly better performances for other variables.The error patterns of the DLMs and OFSs show a high degree of similarity,which is reasonable from the viewpoint of predictability,facilitating further applications of the DLMs.For extreme events,the DLMs and OFSs both present large but similar forecast errors for SLA and current speed,while the DLMs are likely to give larger errors for SST and current direction.This study provides an evaluation of the forecast skills of commonly used DLMs and provides an example to objectively intercompare different DLMs.

Combining deep reinforcement learning with heuristics to solve the traveling salesman problem

Recent studies employing deep learning to solve the traveling salesman problem(TSP)have mainly focused on learning construction heuristics.Such methods can improve TSP solutions,but still depend on additional programs.However,methods that focus on learning improvement heuristics to iteratively refine solutions remain insufficient.Traditional improvement heuristics are guided by a manually designed search strategy and may only achieve limited improvements.This paper proposes a novel framework for learning improvement heuristics,which automatically discovers better improvement policies for heuristics to iteratively solve the TSP.Our framework first designs a new architecture based on a transformer model to make the policy network parameterized,which introduces an action-dropout layer to prevent action selection from overfitting.It then proposes a deep reinforcement learning approach integrating a simulated annealing mechanism(named RL-SA)to learn the pairwise selected policy,aiming to improve the 2-opt algorithm's performance.The RL-SA leverages the whale optimization algorithm to generate initial solutions for better sampling efficiency and uses the Gaussian perturbation strategy to tackle the sparse reward problem of reinforcement learning.The experiment results show that the proposed approach is significantly superior to the state-of-the-art learning-based methods,and further reduces the gap between learning-based methods and highly optimized solvers in the benchmark datasets.Moreover,our pre-trained model M can be applied to guide the SA algorithm(named M-SA(ours)),which performs better than existing deep models in small-,medium-,and large-scale TSPLIB datasets.Additionally,the M-SA(ours)achieves excellent generalization performance in a real-world dataset on global liner shipping routes,with the optimization percentages in distance reduction ranging from3.52%to 17.99%.

ASLP-DL—A Novel Approach Employing Lightweight Deep Learning Framework for Optimizing Accident Severity Level Prediction

Highway safety researchers focus on crash injury severity,utilizing deep learning—specifically,deep neural networks(DNN),deep convolutional neural networks(D-CNN),and deep recurrent neural networks(D-RNN)—as the preferred method for modeling accident severity.Deep learning’s strength lies in handling intricate relation-ships within extensive datasets,making it popular for accident severity level(ASL)prediction and classification.Despite prior success,there is a need for an efficient system recognizing ASL in diverse road conditions.To address this,we present an innovative Accident Severity Level Prediction Deep Learning(ASLP-DL)framework,incorporating DNN,D-CNN,and D-RNN models fine-tuned through iterative hyperparameter selection with Stochastic Gradient Descent.The framework optimizes hidden layers and integrates data augmentation,Gaussian noise,and dropout regularization for improved generalization.Sensitivity and factor contribution analyses identify influential predictors.Evaluated on three diverse crash record databases—NCDB 2018–2019,UK 2015–2020,and US 2016–2021—the D-RNN model excels with an ACC score of 89.0281%,a Roc Area of 0.751,an F-estimate of 0.941,and a Kappa score of 0.0629 over the NCDB dataset.The proposed framework consistently outperforms traditional methods,existing machine learning,and deep learning techniques.

深度学习图像重建算法(DLIR)对能谱CT多参数图像质量改善的体模研究

目的 探讨深度学习重建算法对能谱CT多参数成像单能量图像、有效原子序数图、碘水图、水碘图图像质量的改善。方法 选择一个20cm直径的圆柱形聚丙烯体摸,在内部放入九支试管,试管中分别装入(3.75、7.5、15、30mg/mL)不同浓度及(18m m、10m m、2 m m)不同管径的碘对比剂及水和钙溶液,利用GE APEX CT进行能谱扫描,将扫描完成图像分别利用FBP、40%ASIR-V及深度学习(低DLIR-L、中DLIR-M、高DLIR-H)重建出五组图像。采用后处理软件分别重建出70keV单能量图像、基物质图(碘-水图、水-碘图)及有效原子序数图。对浓度为3.75 mg/m L、15 mg/mL及Water三支试管进行数据分析。在FBP、40%ASIR-V、DLIR-L、DLIR-M、DLIR-H五组图像测量70keV的CT值、碘-水图的碘浓度、水-碘图的水浓度、有效原子序数及图像噪声,计算各图像的信噪比(SNR),对比5组图像质量的差异。结果5种重建算法下的图像在低浓度造影剂(3.75mg/mL)、高浓度造影剂(15mg/m L)及水试管内70keV的CT值差异均无统计学意义(P值均>0.05),有效原子序数、碘水图的碘浓度及水碘图的水浓度亦无明显统计学差异(P值均>0.05)。70keV、有效原子序数图、碘水图、水碘图的噪声及图像信噪比5组图像差异均有统计学意义(P值均<0.05),DLIR下的噪声值均较FBP及40%ASIR-V降低,图像信噪比提高(P值均<0.05),DLIR-H噪声最小,信噪比最高。结论 在能谱CT成像中, DLIR较FBP及40%ASIR-V在单能量图像、有效原子序数图、碘水图、水碘图的噪声降低,信噪比提高。

Deep Learning Hybrid Model for Lithium-Ion Battery Aging Estimation and Prediction

The degradation process of lithium-ion batteries is intricately linked to their entire lifecycle as power sources and energy storage devices,encompassing aspects such as performance delivery and cycling utilization.Consequently,the accurate and expedient estimation or prediction of the aging state of lithium-ion batteries has garnered extensive attention.Nonetheless,prevailing research predominantly concentrates on either aging estimation or prediction,neglecting the dynamic fusion of both facets.This paper proposes a hybrid model for capacity aging estimation and prediction based on deep learning,wherein salient features highly pertinent to aging are extracted from charge and discharge relaxation processes.By amalgamating historical capacity decay data,the model dynamically furnishes estimations of the present capacity and forecasts of future capacity for lithium-ion batteries.Our approach is validated against a novel dataset involving charge and discharge cycles at varying rates.Specifically,under a charging condition of 0.25 C,a mean absolute percentage error(MAPE)of 0.29%is achieved.This outcome underscores the model's adeptness in harnessing relaxation processes commonly encountered in the real world and synergizing with historical capacity records within battery management systems(BMS),thereby affording estimations and prognostications of capacity decline with heightened precision.

基于多尺度特征融合SSDLite的光伏组件缺陷检测

为了应对光伏组件缺陷检测中人工检测速度缓慢以及使用YOLO等深度学习模型时速度较慢且硬件成本高的问题,提出一种基于SSDLite的多层特征融合轻量化目标检测方法。该方法采用MobileNetV2作为SSDLite模型的骨干网络,并从中提取3个不同层次的特征层进行特征融合。针对不同缺陷的尺寸特点,对模型中的先验框的大小也进行了重新设计。在MobileNetV2的瓶颈结构中引入CBAM注意力机制,以提高模型的检测精度。相比传统的SSDLite模型,该文模型平均精度从65.8%提高至72.4%,虽然速度略微下降,但已基本满足实际应用的需求。

Deep learning-based inpainting of saturation artifacts in optical coherence tomography images

Limited by the dynamic range of the detector,saturation artifacts usually occur in optical coherence tomography(OCT)imaging for high scattering media.The available methods are difficult to remove saturation artifacts and restore texture completely in OCT images.We proposed a deep learning-based inpainting method of saturation artifacts in this paper.The generation mechanism of saturation artifacts was analyzed,and experimental and simulated datasets were built based on the mechanism.Enhanced super-resolution generative adversarial networks were trained by the clear–saturated phantom image pairs.The perfect reconstructed results of experimental zebrafish and thyroid OCT images proved its feasibility,strong generalization,and robustness.

Well-defined high entropy-metal nanoparticles:Detection of the multi-element particles by deep learning

Characterizing and control the chemical compositions of multi-element particles as single metal nanoparticles(mNPs) on the surfaces of catalytic metal oxide supports is challenging.This can be attributed to the heterogeneity and large size at the nanoscale,the poorly defined catalyst nanostructure,and thermodynamic immiscibility of the strongly repelling metallic elements.To address these challenges,an ultrasonic-assisted coincident electro-oxidation-reduction-precipitation(U-SEO-P) is presented to fabricate ultra-stable PtRuAgCoCuP NPs,which produces numerous active intermediates and induces strong metal-support interactions.To sort the active high-entropy mNPs,individual NPs are described on the support surface and the role of deep learning in understanding/predicting the features of PtRuAgCoCu@TiO_(x) catalysts is explained.Notably,this deep learning approach required minimal to no human input.The as-prepared PtRuAgCoCu@TiO_(x) catalysts can be used to catalyze various important chemical reactions,such as a high reduction conversion(100% in 30 s),with no loss of catalytic activity even after 20 cycles of nitroarene and ketone/aldehyde,which is several times higher than commercial Pt@TiO_(x) owing to individual PtRuAgCoCuP NPs on TiO_(x) surface.In this study,we present the "Totally Defined Catalysis" concept,which has enormous potential for the advancement of high-activity catalysts in the reduction of organic compounds.

A credibility-aware swarm-federated deep learning framework in internet of vehicles

Although Federated Deep Learning(FDL)enables distributed machine learning in the Internet of Vehicles(IoV),it requires multiple clients to upload model parameters,thus still existing unavoidable communication overhead and data privacy risks.The recently proposed Swarm Learning(SL)provides a decentralized machine learning approach for unit edge computing and blockchain-based coordination.A Swarm-Federated Deep Learning framework in the IoV system(IoV-SFDL)that integrates SL into the FDL framework is proposed in this paper.The IoV-SFDL organizes vehicles to generate local SL models with adjacent vehicles based on the blockchain empowered SL,then aggregates the global FDL model among different SL groups with a credibility weights prediction algorithm.Extensive experimental results show that compared with the baseline frameworks,the proposed IoV-SFDL framework reduces the overhead of client-to-server communication by 16.72%,while the model performance improves by about 5.02%for the same training iterations.

Assessments of Data-Driven Deep Learning Models on One-Month Predictions of Pan-Arctic Sea Ice Thickness

In recent years,deep learning methods have gradually been applied to prediction tasks related to Arctic sea ice concentration,but relatively little research has been conducted for larger spatial and temporal scales,mainly due to the limited time coverage of observations and reanalysis data.Meanwhile,deep learning predictions of sea ice thickness(SIT)have yet to receive ample attention.In this study,two data-driven deep learning(DL)models are built based on the ConvLSTM and fully convolutional U-net(FC-Unet)algorithms and trained using CMIP6 historical simulations for transfer learning and fine-tuned using reanalysis/observations.These models enable monthly predictions of Arctic SIT without considering the complex physical processes involved.Through comprehensive assessments of prediction skills by season and region,the results suggest that using a broader set of CMIP6 data for transfer learning,as well as incorporating multiple climate variables as predictors,contribute to better prediction results,although both DL models can effectively predict the spatiotemporal features of SIT anomalies.Regarding the predicted SIT anomalies of the FC-Unet model,the spatial correlations with reanalysis reach an average level of 89%over all months,while the temporal anomaly correlation coefficients are close to unity in most cases.The models also demonstrate robust performances in predicting SIT and SIE during extreme events.The effectiveness and reliability of the proposed deep transfer learning models in predicting Arctic SIT can facilitate more accurate pan-Arctic predictions,aiding climate change research and real-time business applications.

UAV-Assisted Dynamic Avatar Task Migration for Vehicular Metaverse Services: A Multi-Agent Deep Reinforcement Learning Approach

Avatars, as promising digital representations and service assistants of users in Metaverses, can enable drivers and passengers to immerse themselves in 3D virtual services and spaces of UAV-assisted vehicular Metaverses. However, avatar tasks include a multitude of human-to-avatar and avatar-to-avatar interactive applications, e.g., augmented reality navigation,which consumes intensive computing resources. It is inefficient and impractical for vehicles to process avatar tasks locally. Fortunately, migrating avatar tasks to the nearest roadside units(RSU)or unmanned aerial vehicles(UAV) for execution is a promising solution to decrease computation overhead and reduce task processing latency, while the high mobility of vehicles brings challenges for vehicles to independently perform avatar migration decisions depending on current and future vehicle status. To address these challenges, in this paper, we propose a novel avatar task migration system based on multi-agent deep reinforcement learning(MADRL) to execute immersive vehicular avatar tasks dynamically. Specifically, we first formulate the problem of avatar task migration from vehicles to RSUs/UAVs as a partially observable Markov decision process that can be solved by MADRL algorithms. We then design the multi-agent proximal policy optimization(MAPPO) approach as the MADRL algorithm for the avatar task migration problem. To overcome slow convergence resulting from the curse of dimensionality and non-stationary issues caused by shared parameters in MAPPO, we further propose a transformer-based MAPPO approach via sequential decision-making models for the efficient representation of relationships among agents. Finally, to motivate terrestrial or non-terrestrial edge servers(e.g., RSUs or UAVs) to share computation resources and ensure traceability of the sharing records, we apply smart contracts and blockchain technologies to achieve secure sharing management. Numerical results demonstrate that the proposed approach outperforms the MAPPO approach by around 2% and effectively reduces approximately 20% of the latency of avatar task execution in UAV-assisted vehicular Metaverses.

ST-LSTM-SA:A New Ocean Sound Velocity Field Prediction Model Based on Deep Learning

The scarcity of in-situ ocean observations poses a challenge for real-time information acquisition in the ocean.Among the crucial hydroacoustic environmental parameters,ocean sound velocity exhibits significant spatial and temporal variability and it is highly relevant to oceanic research.In this study,we propose a new data-driven approach,leveraging deep learning techniques,for the prediction of sound velocity fields(SVFs).Our novel spatiotemporal prediction model,STLSTM-SA,combines Spatiotemporal Long Short-Term Memory(ST-LSTM) with a self-attention mechanism to enable accurate and real-time prediction of SVFs.To circumvent the limited amount of observational data,we employ transfer learning by first training the model using reanalysis datasets,followed by fine-tuning it using in-situ analysis data to obtain the final prediction model.By utilizing the historical 12-month SVFs as input,our model predicts the SVFs for the subsequent three months.We compare the performance of five models:Artificial Neural Networks(ANN),Long ShortTerm Memory(LSTM),Convolutional LSTM(ConvLSTM),ST-LSTM,and our proposed ST-LSTM-SA model in a test experiment spanning 2019 to 2022.Our results demonstrate that the ST-LSTM-SA model significantly improves the prediction accuracy and stability of sound velocity in both temporal and spatial dimensions.The ST-LSTM-SA model not only accurately predicts the ocean sound velocity field(SVF),but also provides valuable insights for spatiotemporal prediction of other oceanic environmental variables.