Gait recognition is a key biometric for long-distance identification,yet its performance is severely degraded by real-world challenges such as varying clothing,carrying conditions,and changing viewpoints.While combini...Gait recognition is a key biometric for long-distance identification,yet its performance is severely degraded by real-world challenges such as varying clothing,carrying conditions,and changing viewpoints.While combining silhouette and skeleton data is a promising direction,effectively fusing these heterogeneous modalities and adaptively weighting their contributions in response to diverse conditions remains a central problem.This paper introduces GaitMAFF,a novelMulti-modal Adaptive Feature Fusion Network,to address this challenge.Our approach first transforms discrete skeleton joints into a dense SkeletonMap representation to align with silhouettes,then employs an attention-based module to dynamically learn the fusion weights between the two modalities.These fused features are processed by a powerful spatio-temporal backbone withWeighted Global-Local Feature FusionModules(WFFM)to learn a discriminative representation.Extensive experiments on the challenging CCPG and Gait3D datasets show that GaitMAFF achieves state-of-the-art performance,with an average Rank-1 accuracy of 84.6%on CCPG and 58.7%on Gait3D.These results demonstrate that our adaptive fusion strategy effectively integrates complementary multimodal information,significantly enhancing gait recognition robustness and accuracy in complex scenes and providing a practical solution for real-world applications.展开更多
Background Computed tomography(CT) and cone-beam computed tomography(CBCT) image registration play pivotal roles in computer-assisted navigation for orthopedic surgery. Traditional methods often apply uniform deformat...Background Computed tomography(CT) and cone-beam computed tomography(CBCT) image registration play pivotal roles in computer-assisted navigation for orthopedic surgery. Traditional methods often apply uniform deformation models, neglecting the biomechanical differences between rigid structures and soft tissues, which compromises registration accuracy, especially during significant bone displacements. Method To address this issue, we introduce RE-Reg, a rigid-elastic CT-CBCT image registration framework that jointly learns rigid bone motion and soft tissue deformation. RE-Reg incorporates a rigid alignment(RA) module to estimate global bone motion and an elastic deformation(ED) module to model soft tissue deformation, preserving bony structures through bone shape preservation(BSP) loss. Result Our comprehensive evaluation on publicly available datasets demonstrates that RE-Reg significantly outperforms existing methods in terms of registration accuracy and rigid bone structure preservation, achieving a 1.3% improvement in Dice similarity coefficient(DSC) and a 23% reduction in rigid bone deformation(%Δvol) compared with the best baseline. Conclusion This framework not only enhances anatomical fidelity but also ensures biomechanical plausibility and provides a valuable tool for image-guided orthopedic surgery. This code is available athttps://github.com/Zq-Huang/RE-Reg.展开更多
The fasteners employed in the railway tracks are susceptible to defects arising from their intricate composition.Foreign objects are frequently observed on the track bed in an open environment.These two types of defec...The fasteners employed in the railway tracks are susceptible to defects arising from their intricate composition.Foreign objects are frequently observed on the track bed in an open environment.These two types of defects pose potential threats to high-speed trains,thus necessitating timely and accurate track inspection.The majority of extant automatic inspection methods are predicated on the utilization of single visible light data,and the efficacy of the algorithmic processes is influenced by complex environments.Furthermore,due to the single information dimension,the detection accuracy of defects in similar,occluded,and small object categories is low.To address the aforementioned issues,this paper proposes a track defect detectionmethod based on dynamicmulti-modal fusion and challenging object enhanced perception.First,in light of the variances in the representation dimensions ofmultimodal information,this paper proposes a dynamic weighted multi-modal feature fusion module.The fused multi-modal features are assigned weights,and thenmultiplied with the extracted single-modal features atmultiple levels,achieving adaptive adjustment of the response degree of fusion features.Second,a novel stepwise multi-scale convolution feature aggregation module is proposed for challenging objects.The proposed method employs depth separable convolution and cross-scale aggregation operations of different receptive fields to enhance feature extraction and reuse,thereby reducing the degree of progressive loss of effective information.The experimental results demonstrate the efficacy of the proposed method in comparison to eight established methods,encompassing both single-modal and multi-modal methods,as evidenced by the extensive findings within the constructed RGBD dataset.展开更多
Autism spectrum disorder(AsD)is a highly heterogeneous neurodevelopmental disorder.Early diagnosis and intervention are crucial for improving outcomes.Traditional single-modality diagnostic methods are subjective,limi...Autism spectrum disorder(AsD)is a highly heterogeneous neurodevelopmental disorder.Early diagnosis and intervention are crucial for improving outcomes.Traditional single-modality diagnostic methods are subjective,limited,and struggle to reveal the underlying pathological mechanisms.In contrast,multimodal data analysis integrates behavioral,physiological,and neuroimaging information with advanced machine-learning and deeplearning algorithms to overcome these limitations.In this review,we surveyed the recent pediatric AsD literature,highlighting artificial intelligence-driven diagnostic techniques,multimodal data fusion strategies,and emerging trends in ASD assessment.We surveyed studies that integrated two or more modalities and summarized the fusion levels,learning paradigms,tasks,datasets,and metrics.Multimodal approaches outperform singlemodality baselines in classification,severity estimation,and subtyping by leveraging complementary information and reducing modality-specific biases.Multimodal approaches significantly enhance diagnostic accuracy and comprehensiveness,enabling early screening of AsD,symptom subtyping,severity assessment,and personalized interventions.Advances in multimodal fusion techniques have promoted progress in precision medicine for the treatment of ASD.展开更多
Image registration within a solar photosphere sequence is crucial for observational solar physics studies requiring high spatial and temporal resolutions.Previously,we identified residual large-scale nonrigid distorti...Image registration within a solar photosphere sequence is crucial for observational solar physics studies requiring high spatial and temporal resolutions.Previously,we identified residual large-scale nonrigid distortions in high-resolution solar photosphere images from ground-based telescopes after high-resolution reconstruction.Because these distortions are not eliminated by conventional sequence correlation alignment,they can affect the analysis of small-scale activity in the solar photosphere.Here,we implemented an image registration model using deep learning(HCAM-Net)to solve the problem.Within an encoder-decoder framework,we introduced a hybrid attention mechanism to improve context information capture and extract accurate deformation fields.Analyzing solar photosphere images acquired by the New Vacuum Solar Telescope,we demonstrated that the proposed model effectively achieved highly accurate nonrigid image registration.Evaluation metrics and visualization results indicated that our model outperformed current state-of-the-art models,such as VoxelMorph and TransMorph,for nonrigid registration of solar photosphere images,with a structural similarity index measure of 0.965 and a coefficient of determination of 0.976.展开更多
In multi-modal emotion recognition,excessive reliance on historical context often impedes the detection of emotional shifts,while modality heterogeneity and unimodal noise limit recognition performance.Existing method...In multi-modal emotion recognition,excessive reliance on historical context often impedes the detection of emotional shifts,while modality heterogeneity and unimodal noise limit recognition performance.Existing methods struggle to dynamically adjust cross-modal complementary strength to optimize fusion quality and lack effective mechanisms to model the dynamic evolution of emotions.To address these issues,we propose a multi-level dynamic gating and emotion transfer framework for multi-modal emotion recognition.A dynamic gating mechanism is applied across unimodal encoding,cross-modal alignment,and emotion transfer modeling,substantially improving noise robustness and feature alignment.First,we construct a unimodal encoder based on gated recurrent units and feature-selection gating to suppress intra-modal noise and enhance contextual representation.Second,we design a gated-attention crossmodal encoder that dynamically calibrates the complementary contributions of visual and audio modalities to the dominant textual features and eliminates redundant information.Finally,we introduce a gated enhanced emotion transfer module that explicitly models the temporal dependence of emotional evolution in dialogues via transfer gating and optimizes continuity modeling with a comparative learning loss.Experimental results demonstrate that the proposed method outperforms state-of-the-art models on the public MELD and IEMOCAP datasets.展开更多
To address the challenge of achieving decentralized,scalable,and adaptive control for large-scale multiple unmanned aerial vehicle(multi-UAV)swarms in dynamic urban environments with obstacles and wind perturbations,w...To address the challenge of achieving decentralized,scalable,and adaptive control for large-scale multiple unmanned aerial vehicle(multi-UAV)swarms in dynamic urban environments with obstacles and wind perturbations,we proposed a hybrid framework integrating adaptive reinforcement learning(RL),multi-modal perception fusion,and enhanced pigeon flock optimization(PFO)with curiosity-driven exploration to enable robust autonomous and formation control.The framework leverages meta-learning to optimize RL policies for real-time adaptation,fuses sensor data for precise state estimation,and enhances PFO with learned leader-follower dynamics and exploration rewards to maintain cohesive formations and explore uncertain areas.For swarms of 10–30 UAVs,it achieves 34%faster convergence,61%reduced stability root mean square error(RMSE),88%fewer collisions and 85.6%–92.3%success rates in target detection and encirclement,outperforming standard multi-agent RL,pure PFO,and single-modality RL.Three-dimensional trajectory visualizations confirm cohesive formations,collision-free maneuvers,and efficient exploration in urban search-and-rescue scenarios.Innovations include meta-RL for rapid adaptation,multi-modal fusion for robust perception,and curiosity-driven PFO for scalable,decentralized control,advancing real-world multi-UAV swarm autonomy and coordination.展开更多
A new coarse-to-fine strategy was proposed for nonrigid registration of computed tomography(CT) and magnetic resonance(MR) images of a liver.This hierarchical framework consisted of an affine transformation and a B-sp...A new coarse-to-fine strategy was proposed for nonrigid registration of computed tomography(CT) and magnetic resonance(MR) images of a liver.This hierarchical framework consisted of an affine transformation and a B-splines free-form deformation(FFD).The affine transformation performed a rough registration targeting the mismatch between the CT and MR images.The B-splines FFD transformation performed a finer registration by correcting local motion deformation.In the registration algorithm,the normalized mutual information(NMI) was used as similarity measure,and the limited memory Broyden-Fletcher- Goldfarb-Shannon(L-BFGS) optimization method was applied for optimization process.The algorithm was applied to the fully automated registration of liver CT and MR images in three subjects.The results demonstrate that the proposed method not only significantly improves the registration accuracy but also reduces the running time,which is effective and efficient for nonrigid registration.展开更多
In order to overcome the defects where the surface of the object lacks sufficient texture features and the algorithm cannot meet the real-time requirements of augmented reality,a markerless augmented reality tracking ...In order to overcome the defects where the surface of the object lacks sufficient texture features and the algorithm cannot meet the real-time requirements of augmented reality,a markerless augmented reality tracking registration method based on multimodal template matching and point clouds is proposed.The method first adapts the linear parallel multi-modal LineMod template matching method with scale invariance to identify the texture-less target and obtain the reference image as the key frame that is most similar to the current perspective.Then,we can obtain the initial pose of the camera and solve the problem of re-initialization because of tracking registration interruption.A point cloud-based method is used to calculate the precise pose of the camera in real time.In order to solve the problem that the traditional iterative closest point(ICP)algorithm cannot meet the real-time requirements of the system,Kdtree(k-dimensional tree)is used under the graphics processing unit(GPU)to replace the part of finding the nearest points in the original ICP algorithm to improve the speed of tracking registration.At the same time,the random sample consensus(RANSAC)algorithm is used to remove the error point pairs to improve the accuracy of the algorithm.The results show that the proposed tracking registration method has good real-time performance and robustness.展开更多
In this work, we propose a new variational model for multi-modal image registration and present an efficient numerical implementation. The model minimizes a new functional based on using reformulated normalized gradie...In this work, we propose a new variational model for multi-modal image registration and present an efficient numerical implementation. The model minimizes a new functional based on using reformulated normalized gradients of the images as the fidelity term and higher-order derivatives as the regularizer. A key feature of the model is its ability of guaranteeing a diffeomorphic transformation which is achieved by a control term motivated by the quasi-conformal map and Beltrami coefficient. The existence of the solution of this model is established. To solve the model numerically, we design a Gauss-Newton method to solve the resulting discrete optimization problem and prove its convergence;a multilevel technique is employed to speed up the initialization and avoid likely local minima of the underlying functional. Finally, numerical experiments demonstrate that this new model can deliver good performances for multi-modal image registration and simultaneously generate an accurate diffeomorphic transformation.展开更多
This paper describes a new method for three-dimensional medical image registration. In the interactive image-guided HIFU （ High Intensity Focused Ultrasound） therapy system, a fast and precise localization of the tu...This paper describes a new method for three-dimensional medical image registration. In the interactive image-guided HIFU （ High Intensity Focused Ultrasound） therapy system, a fast and precise localization of the tumor is very important. An automatic system is developed for registering pre-operative MR images with intra-operative ultrasound images based on the vessels visible in both of the modalities. When the MR and the ultrasound images are aligned, the eenterline points of the vessels in the MR image will align with bright intensities in the ultrasound image. The method applies an optimization strategy combining the genetic algorithm with the conjugated gradients algorithm to minimize the objective function. It provides a feasible way of determining the global solution and makes the method robust to local maximum and insensitive to initial position. Two experiments were designed to evaluate the method, and the results show that our method has better registration accuracy and convergence rate than the other two classic algorithms.展开更多
Objective To develop a non-invasive predictive model for coronary artery stenosis severity based on adaptive multi-modal integration of traditional Chinese and western medicine data.Methods Clinical indicators,echocar...Objective To develop a non-invasive predictive model for coronary artery stenosis severity based on adaptive multi-modal integration of traditional Chinese and western medicine data.Methods Clinical indicators,echocardiographic data,traditional Chinese medicine(TCM)tongue manifestations,and facial features were collected from patients who underwent coro-nary computed tomography angiography(CTA)in the Cardiac Care Unit(CCU)of Shanghai Tenth People's Hospital between May 1,2023 and May 1,2024.An adaptive weighted multi-modal data fusion(AWMDF)model based on deep learning was constructed to predict the severity of coronary artery stenosis.The model was evaluated using metrics including accura-cy,precision,recall,F1 score,and the area under the receiver operating characteristic(ROC)curve(AUC).Further performance assessment was conducted through comparisons with six ensemble machine learning methods,data ablation,model component ablation,and various decision-level fusion strategies.Results A total of 158 patients were included in the study.The AWMDF model achieved ex-cellent predictive performance(AUC=0.973,accuracy=0.937,precision=0.937,recall=0.929,and F1 score=0.933).Compared with model ablation,data ablation experiments,and various traditional machine learning models,the AWMDF model demonstrated superior per-formance.Moreover,the adaptive weighting strategy outperformed alternative approaches,including simple weighting,averaging,voting,and fixed-weight schemes.Conclusion The AWMDF model demonstrates potential clinical value in the non-invasive prediction of coronary artery disease and could serve as a tool for clinical decision support.展开更多
Traditional Chinese medicine(TCM)demonstrates distinctive advantages in disease prevention and treatment.However,analyzing its biological mechanisms through the modern medical research paradigm of“single drug,single ...Traditional Chinese medicine(TCM)demonstrates distinctive advantages in disease prevention and treatment.However,analyzing its biological mechanisms through the modern medical research paradigm of“single drug,single target”presents significant challenges due to its holistic approach.Network pharmacology and its core theory of network targets connect drugs and diseases from a holistic and systematic perspective based on biological networks,overcoming the limitations of reductionist research models and showing considerable value in TCM research.Recent integration of network target computational and experimental methods with artificial intelligence(AI)and multi-modal multi-omics technologies has substantially enhanced network pharmacology methodology.The advancement in computational and experimental techniques provides complementary support for network target theory in decoding TCM principles.This review,centered on network targets,examines the progress of network target methods combined with AI in predicting disease molecular mechanisms and drug-target relationships,alongside the application of multi-modal multi-omics technologies in analyzing TCM formulae,syndromes,and toxicity.Looking forward,network target theory is expected to incorporate emerging technologies while developing novel approaches aligned with its unique characteristics,potentially leading to significant breakthroughs in TCM research and advancing scientific understanding and innovation in TCM.展开更多
The multi-modal characteristics of mineral particles play a pivotal role in enhancing the classification accuracy,which is critical for obtaining a profound understanding of the Earth's composition and ensuring ef...The multi-modal characteristics of mineral particles play a pivotal role in enhancing the classification accuracy,which is critical for obtaining a profound understanding of the Earth's composition and ensuring effective exploitation utilization of its resources.However,the existing methods for classifying mineral particles do not fully utilize these multi-modal features,thereby limiting the classification accuracy.Furthermore,when conventional multi-modal image classification methods are applied to planepolarized and cross-polarized sequence images of mineral particles,they encounter issues such as information loss,misaligned features,and challenges in spatiotemporal feature extraction.To address these challenges,we propose a multi-modal mineral particle polarization image classification network(MMGC-Net)for precise mineral particle classification.Initially,MMGC-Net employs a two-dimensional(2D)backbone network with shared parameters to extract features from two types of polarized images to ensure feature alignment.Subsequently,a cross-polarized intra-modal feature fusion module is designed to refine the spatiotemporal features from the extracted features of the cross-polarized sequence images.Ultimately,the inter-modal feature fusion module integrates the two types of modal features to enhance the classification precision.Quantitative and qualitative experimental results indicate that when compared with the current state-of-the-art multi-modal image classification methods,MMGC-Net demonstrates marked superiority in terms of mineral particle multi-modal feature learning and four classification evaluation metrics.It also demonstrates better stability than the existing models.展开更多
With the advent of the next-generation Air Traffic Control(ATC)system,there is growing interest in using Artificial Intelligence(AI)techniques to enhance Situation Awareness(SA)for ATC Controllers(ATCOs),i.e.,Intellig...With the advent of the next-generation Air Traffic Control(ATC)system,there is growing interest in using Artificial Intelligence(AI)techniques to enhance Situation Awareness(SA)for ATC Controllers(ATCOs),i.e.,Intelligent SA(ISA).However,the existing AI-based SA approaches often rely on unimodal data and lack a comprehensive description and benchmark of the ISA tasks utilizing multi-modal data for real-time ATC environments.To address this gap,by analyzing the situation awareness procedure of the ATCOs,the ISA task is refined to the processing of the two primary elements,i.e.,spoken instructions and flight trajectories.Subsequently,the ISA is further formulated into Controlling Intent Understanding(CIU)and Flight Trajectory Prediction(FTP)tasks.For the CIU task,an innovative automatic speech recognition and understanding framework is designed to extract the controlling intent from unstructured and continuous ATC communications.For the FTP task,the single-and multi-horizon FTP approaches are investigated to support the high-precision prediction of the situation evolution.A total of 32 unimodal/multi-modal advanced methods with extensive evaluation metrics are introduced to conduct the benchmarks on the real-world multi-modal ATC situation dataset.Experimental results demonstrate the effectiveness of AI-based techniques in enhancing ISA for the ATC environment.展开更多
A personalized outfit recommendation has emerged as a hot research topic in the fashion domain.However,existing recommendations do not fully exploit user style preferences.Typically,users prefer particular styles such...A personalized outfit recommendation has emerged as a hot research topic in the fashion domain.However,existing recommendations do not fully exploit user style preferences.Typically,users prefer particular styles such as casual and athletic styles,and consider attributes like color and texture when selecting outfits.To achieve personalized outfit recommendations in line with user style preferences,this paper proposes a personal style guided outfit recommendation with multi-modal fashion compatibility modeling,termed as PSGNet.Firstly,a style classifier is designed to categorize fashion images of various clothing types and attributes into distinct style categories.Secondly,a personal style prediction module extracts user style preferences by analyzing historical data.Then,to address the limitations of single-modal representations and enhance fashion compatibility,both fashion images and text data are leveraged to extract multi-modal features.Finally,PSGNet integrates these components through Bayesian personalized ranking(BPR)to unify the personal style and fashion compatibility,where the former is used as personal style features and guides the output of the personalized outfit recommendation tailored to the target user.Extensive experiments on large-scale datasets demonstrate that the proposed model is efficient on the personalized outfit recommendation.展开更多
With the rapid progress of the artificial intelligence(AI)technology and mobile internet,3D hand pose estimation has become critical to various intelligent application areas,e.g.,human-computer interaction.To avoid th...With the rapid progress of the artificial intelligence(AI)technology and mobile internet,3D hand pose estimation has become critical to various intelligent application areas,e.g.,human-computer interaction.To avoid the low accuracy of single-modal estimation and the high complexity of traditional multi-modal 3D estimation,this paper proposes a novel multi-modal multi-view(MMV)3D hand pose estimation system,which introduces a registration before translation(RT)-translation before registration(TR)jointed conditional generative adversarial network(cGAN)to train a multi-modal registration network,and then employs the multi-modal feature fusion to achieve high-quality estimation,with low hardware and software costs both in data acquisition and processing.Experimental results demonstrate that the MMV system is effective and feasible in various scenarios.It is promising for the MMV system to be used in broad intelligent application areas.展开更多
Multi-modal Named Entity Recognition(MNER)aims to better identify meaningful textual entities by integrating information from images.Previous work has focused on extracting visual semantics at a fine-grained level,or ...Multi-modal Named Entity Recognition(MNER)aims to better identify meaningful textual entities by integrating information from images.Previous work has focused on extracting visual semantics at a fine-grained level,or obtaining entity related external knowledge from knowledge bases or Large Language Models(LLMs).However,these approaches ignore the poor semantic correlation between visual and textual modalities in MNER datasets and do not explore different multi-modal fusion approaches.In this paper,we present MMAVK,a multi-modal named entity recognition model with auxiliary visual knowledge and word-level fusion,which aims to leverage the Multi-modal Large Language Model(MLLM)as an implicit knowledge base.It also extracts vision-based auxiliary knowledge from the image formore accurate and effective recognition.Specifically,we propose vision-based auxiliary knowledge generation,which guides the MLLM to extract external knowledge exclusively derived from images to aid entity recognition by designing target-specific prompts,thus avoiding redundant recognition and cognitive confusion caused by the simultaneous processing of image-text pairs.Furthermore,we employ a word-level multi-modal fusion mechanism to fuse the extracted external knowledge with each word-embedding embedded from the transformerbased encoder.Extensive experimental results demonstrate that MMAVK outperforms or equals the state-of-the-art methods on the two classical MNER datasets,even when the largemodels employed have significantly fewer parameters than other baselines.展开更多
Multi-modal knowledge graph completion(MMKGC)aims to complete missing entities or relations in multi-modal knowledge graphs,thereby discovering more previously unknown triples.Due to the continuous growth of data and ...Multi-modal knowledge graph completion(MMKGC)aims to complete missing entities or relations in multi-modal knowledge graphs,thereby discovering more previously unknown triples.Due to the continuous growth of data and knowledge and the limitations of data sources,the visual knowledge within the knowledge graphs is generally of low quality,and some entities suffer from the issue of missing visual modality.Nevertheless,previous studies of MMKGC have primarily focused on how to facilitate modality interaction and fusion while neglecting the problems of low modality quality and modality missing.In this case,mainstream MMKGC models only use pre-trained visual encoders to extract features and transfer the semantic information to the joint embeddings through modal fusion,which inevitably suffers from problems such as error propagation and increased uncertainty.To address these problems,we propose a Multi-modal knowledge graph Completion model based on Super-resolution and Detailed Description Generation(MMCSD).Specifically,we leverage a pre-trained residual network to enhance the resolution and improve the quality of the visual modality.Moreover,we design multi-level visual semantic extraction and entity description generation,thereby further extracting entity semantics from structural triples and visual images.Meanwhile,we train a variational multi-modal auto-encoder and utilize a pre-trained multi-modal language model to complement the missing visual features.We conducted experiments on FB15K-237 and DB13K,and the results showed that MMCSD can effectively perform MMKGC and achieve state-of-the-art performance.展开更多
Integrating multiple medical imaging techniques,including Magnetic Resonance Imaging(MRI),Computed Tomography,Positron Emission Tomography(PET),and ultrasound,provides a comprehensive view of the patient health status...Integrating multiple medical imaging techniques,including Magnetic Resonance Imaging(MRI),Computed Tomography,Positron Emission Tomography(PET),and ultrasound,provides a comprehensive view of the patient health status.Each of these methods contributes unique diagnostic insights,enhancing the overall assessment of patient condition.Nevertheless,the amalgamation of data from multiple modalities presents difficulties due to disparities in resolution,data collection methods,and noise levels.While traditional models like Convolutional Neural Networks(CNNs)excel in single-modality tasks,they struggle to handle multi-modal complexities,lacking the capacity to model global relationships.This research presents a novel approach for examining multi-modal medical imagery using a transformer-based system.The framework employs self-attention and cross-attention mechanisms to synchronize and integrate features across various modalities.Additionally,it shows resilience to variations in noise and image quality,making it adaptable for real-time clinical use.To address the computational hurdles linked to transformer models,particularly in real-time clinical applications in resource-constrained environments,several optimization techniques have been integrated to boost scalability and efficiency.Initially,a streamlined transformer architecture was adopted to minimize the computational load while maintaining model effectiveness.Methods such as model pruning,quantization,and knowledge distillation have been applied to reduce the parameter count and enhance the inference speed.Furthermore,efficient attention mechanisms such as linear or sparse attention were employed to alleviate the substantial memory and processing requirements of traditional self-attention operations.For further deployment optimization,researchers have implemented hardware-aware acceleration strategies,including the use of TensorRT and ONNX-based model compression,to ensure efficient execution on edge devices.These optimizations allow the approach to function effectively in real-time clinical settings,ensuring viability even in environments with limited resources.Future research directions include integrating non-imaging data to facilitate personalized treatment and enhancing computational efficiency for implementation in resource-limited environments.This study highlights the transformative potential of transformer models in multi-modal medical imaging,offering improvements in diagnostic accuracy and patient care outcomes.展开更多
基金funded by the Natural Science Foundation of Chongqing Municipality,grant number CSTB2022NSCQ-MSX0503.
文摘Gait recognition is a key biometric for long-distance identification,yet its performance is severely degraded by real-world challenges such as varying clothing,carrying conditions,and changing viewpoints.While combining silhouette and skeleton data is a promising direction,effectively fusing these heterogeneous modalities and adaptively weighting their contributions in response to diverse conditions remains a central problem.This paper introduces GaitMAFF,a novelMulti-modal Adaptive Feature Fusion Network,to address this challenge.Our approach first transforms discrete skeleton joints into a dense SkeletonMap representation to align with silhouettes,then employs an attention-based module to dynamically learn the fusion weights between the two modalities.These fused features are processed by a powerful spatio-temporal backbone withWeighted Global-Local Feature FusionModules(WFFM)to learn a discriminative representation.Extensive experiments on the challenging CCPG and Gait3D datasets show that GaitMAFF achieves state-of-the-art performance,with an average Rank-1 accuracy of 84.6%on CCPG and 58.7%on Gait3D.These results demonstrate that our adaptive fusion strategy effectively integrates complementary multimodal information,significantly enhancing gait recognition robustness and accuracy in complex scenes and providing a practical solution for real-world applications.
基金Supported by the National Natural Science Foundation of China(Grant Nos.62025104,62331005,and U22A2052)the Beijing Natural Science Foundation(Grant No.L242100).
文摘Background Computed tomography(CT) and cone-beam computed tomography(CBCT) image registration play pivotal roles in computer-assisted navigation for orthopedic surgery. Traditional methods often apply uniform deformation models, neglecting the biomechanical differences between rigid structures and soft tissues, which compromises registration accuracy, especially during significant bone displacements. Method To address this issue, we introduce RE-Reg, a rigid-elastic CT-CBCT image registration framework that jointly learns rigid bone motion and soft tissue deformation. RE-Reg incorporates a rigid alignment(RA) module to estimate global bone motion and an elastic deformation(ED) module to model soft tissue deformation, preserving bony structures through bone shape preservation(BSP) loss. Result Our comprehensive evaluation on publicly available datasets demonstrates that RE-Reg significantly outperforms existing methods in terms of registration accuracy and rigid bone structure preservation, achieving a 1.3% improvement in Dice similarity coefficient(DSC) and a 23% reduction in rigid bone deformation(%Δvol) compared with the best baseline. Conclusion This framework not only enhances anatomical fidelity but also ensures biomechanical plausibility and provides a valuable tool for image-guided orthopedic surgery. This code is available athttps://github.com/Zq-Huang/RE-Reg.
基金funded by Beijing Natural Science Foundation,grant number L241078.
文摘The fasteners employed in the railway tracks are susceptible to defects arising from their intricate composition.Foreign objects are frequently observed on the track bed in an open environment.These two types of defects pose potential threats to high-speed trains,thus necessitating timely and accurate track inspection.The majority of extant automatic inspection methods are predicated on the utilization of single visible light data,and the efficacy of the algorithmic processes is influenced by complex environments.Furthermore,due to the single information dimension,the detection accuracy of defects in similar,occluded,and small object categories is low.To address the aforementioned issues,this paper proposes a track defect detectionmethod based on dynamicmulti-modal fusion and challenging object enhanced perception.First,in light of the variances in the representation dimensions ofmultimodal information,this paper proposes a dynamic weighted multi-modal feature fusion module.The fused multi-modal features are assigned weights,and thenmultiplied with the extracted single-modal features atmultiple levels,achieving adaptive adjustment of the response degree of fusion features.Second,a novel stepwise multi-scale convolution feature aggregation module is proposed for challenging objects.The proposed method employs depth separable convolution and cross-scale aggregation operations of different receptive fields to enhance feature extraction and reuse,thereby reducing the degree of progressive loss of effective information.The experimental results demonstrate the efficacy of the proposed method in comparison to eight established methods,encompassing both single-modal and multi-modal methods,as evidenced by the extensive findings within the constructed RGBD dataset.
基金supported by the National Key Research and Development Program of China(Research Grant Number:2023YFC3603600).
文摘Autism spectrum disorder(AsD)is a highly heterogeneous neurodevelopmental disorder.Early diagnosis and intervention are crucial for improving outcomes.Traditional single-modality diagnostic methods are subjective,limited,and struggle to reveal the underlying pathological mechanisms.In contrast,multimodal data analysis integrates behavioral,physiological,and neuroimaging information with advanced machine-learning and deeplearning algorithms to overcome these limitations.In this review,we surveyed the recent pediatric AsD literature,highlighting artificial intelligence-driven diagnostic techniques,multimodal data fusion strategies,and emerging trends in ASD assessment.We surveyed studies that integrated two or more modalities and summarized the fusion levels,learning paradigms,tasks,datasets,and metrics.Multimodal approaches outperform singlemodality baselines in classification,severity estimation,and subtyping by leveraging complementary information and reducing modality-specific biases.Multimodal approaches significantly enhance diagnostic accuracy and comprehensiveness,enabling early screening of AsD,symptom subtyping,severity assessment,and personalized interventions.Advances in multimodal fusion techniques have promoted progress in precision medicine for the treatment of ASD.
基金funded by the Strategic Priority Research Program of the Chinese Academy of Sciences(XDB0560000)the National Natural Science Foundation of China(12473054)+1 种基金the Basic Research on Fund Projects in Yunnan Province(2019FA001)the Yunnan Province Science Foundation Project(202105AC160085).
文摘Image registration within a solar photosphere sequence is crucial for observational solar physics studies requiring high spatial and temporal resolutions.Previously,we identified residual large-scale nonrigid distortions in high-resolution solar photosphere images from ground-based telescopes after high-resolution reconstruction.Because these distortions are not eliminated by conventional sequence correlation alignment,they can affect the analysis of small-scale activity in the solar photosphere.Here,we implemented an image registration model using deep learning(HCAM-Net)to solve the problem.Within an encoder-decoder framework,we introduced a hybrid attention mechanism to improve context information capture and extract accurate deformation fields.Analyzing solar photosphere images acquired by the New Vacuum Solar Telescope,we demonstrated that the proposed model effectively achieved highly accurate nonrigid image registration.Evaluation metrics and visualization results indicated that our model outperformed current state-of-the-art models,such as VoxelMorph and TransMorph,for nonrigid registration of solar photosphere images,with a structural similarity index measure of 0.965 and a coefficient of determination of 0.976.
基金funded by“the Fanying Special Program of the National Natural Science Foundation of China,grant number 62341307”“the Scientific research project of Jiangxi Provincial Department of Education,grant number GJJ200839”“theDoctoral startup fund of JiangxiUniversity of Technology,grant number 205200100402”.
文摘In multi-modal emotion recognition,excessive reliance on historical context often impedes the detection of emotional shifts,while modality heterogeneity and unimodal noise limit recognition performance.Existing methods struggle to dynamically adjust cross-modal complementary strength to optimize fusion quality and lack effective mechanisms to model the dynamic evolution of emotions.To address these issues,we propose a multi-level dynamic gating and emotion transfer framework for multi-modal emotion recognition.A dynamic gating mechanism is applied across unimodal encoding,cross-modal alignment,and emotion transfer modeling,substantially improving noise robustness and feature alignment.First,we construct a unimodal encoder based on gated recurrent units and feature-selection gating to suppress intra-modal noise and enhance contextual representation.Second,we design a gated-attention crossmodal encoder that dynamically calibrates the complementary contributions of visual and audio modalities to the dominant textual features and eliminates redundant information.Finally,we introduce a gated enhanced emotion transfer module that explicitly models the temporal dependence of emotional evolution in dialogues via transfer gating and optimizes continuity modeling with a comparative learning loss.Experimental results demonstrate that the proposed method outperforms state-of-the-art models on the public MELD and IEMOCAP datasets.
基金supported by the National Natural Science Foundation of China(No.62350048)。
文摘To address the challenge of achieving decentralized,scalable,and adaptive control for large-scale multiple unmanned aerial vehicle(multi-UAV)swarms in dynamic urban environments with obstacles and wind perturbations,we proposed a hybrid framework integrating adaptive reinforcement learning(RL),multi-modal perception fusion,and enhanced pigeon flock optimization(PFO)with curiosity-driven exploration to enable robust autonomous and formation control.The framework leverages meta-learning to optimize RL policies for real-time adaptation,fuses sensor data for precise state estimation,and enhances PFO with learned leader-follower dynamics and exploration rewards to maintain cohesive formations and explore uncertain areas.For swarms of 10–30 UAVs,it achieves 34%faster convergence,61%reduced stability root mean square error(RMSE),88%fewer collisions and 85.6%–92.3%success rates in target detection and encirclement,outperforming standard multi-agent RL,pure PFO,and single-modality RL.Three-dimensional trajectory visualizations confirm cohesive formations,collision-free maneuvers,and efficient exploration in urban search-and-rescue scenarios.Innovations include meta-RL for rapid adaptation,multi-modal fusion for robust perception,and curiosity-driven PFO for scalable,decentralized control,advancing real-world multi-UAV swarm autonomy and coordination.
基金Project(61240010)supported by the National Natural Science Foundation of ChinaProject(20070007070)supported by Specialized Research Fund for the Doctoral Program of Higher Education of China
文摘A new coarse-to-fine strategy was proposed for nonrigid registration of computed tomography(CT) and magnetic resonance(MR) images of a liver.This hierarchical framework consisted of an affine transformation and a B-splines free-form deformation(FFD).The affine transformation performed a rough registration targeting the mismatch between the CT and MR images.The B-splines FFD transformation performed a finer registration by correcting local motion deformation.In the registration algorithm,the normalized mutual information(NMI) was used as similarity measure,and the limited memory Broyden-Fletcher- Goldfarb-Shannon(L-BFGS) optimization method was applied for optimization process.The algorithm was applied to the fully automated registration of liver CT and MR images in three subjects.The results demonstrate that the proposed method not only significantly improves the registration accuracy but also reduces the running time,which is effective and efficient for nonrigid registration.
基金This work was supported by National Natural Science Foundation of China(No.61125101).
文摘In order to overcome the defects where the surface of the object lacks sufficient texture features and the algorithm cannot meet the real-time requirements of augmented reality,a markerless augmented reality tracking registration method based on multimodal template matching and point clouds is proposed.The method first adapts the linear parallel multi-modal LineMod template matching method with scale invariance to identify the texture-less target and obtain the reference image as the key frame that is most similar to the current perspective.Then,we can obtain the initial pose of the camera and solve the problem of re-initialization because of tracking registration interruption.A point cloud-based method is used to calculate the precise pose of the camera in real time.In order to solve the problem that the traditional iterative closest point(ICP)algorithm cannot meet the real-time requirements of the system,Kdtree(k-dimensional tree)is used under the graphics processing unit(GPU)to replace the part of finding the nearest points in the original ICP algorithm to improve the speed of tracking registration.At the same time,the random sample consensus(RANSAC)algorithm is used to remove the error point pairs to improve the accuracy of the algorithm.The results show that the proposed tracking registration method has good real-time performance and robustness.
文摘In this work, we propose a new variational model for multi-modal image registration and present an efficient numerical implementation. The model minimizes a new functional based on using reformulated normalized gradients of the images as the fidelity term and higher-order derivatives as the regularizer. A key feature of the model is its ability of guaranteeing a diffeomorphic transformation which is achieved by a control term motivated by the quasi-conformal map and Beltrami coefficient. The existence of the solution of this model is established. To solve the model numerically, we design a Gauss-Newton method to solve the resulting discrete optimization problem and prove its convergence;a multilevel technique is employed to speed up the initialization and avoid likely local minima of the underlying functional. Finally, numerical experiments demonstrate that this new model can deliver good performances for multi-modal image registration and simultaneously generate an accurate diffeomorphic transformation.
文摘This paper describes a new method for three-dimensional medical image registration. In the interactive image-guided HIFU （ High Intensity Focused Ultrasound） therapy system, a fast and precise localization of the tumor is very important. An automatic system is developed for registering pre-operative MR images with intra-operative ultrasound images based on the vessels visible in both of the modalities. When the MR and the ultrasound images are aligned, the eenterline points of the vessels in the MR image will align with bright intensities in the ultrasound image. The method applies an optimization strategy combining the genetic algorithm with the conjugated gradients algorithm to minimize the objective function. It provides a feasible way of determining the global solution and makes the method robust to local maximum and insensitive to initial position. Two experiments were designed to evaluate the method, and the results show that our method has better registration accuracy and convergence rate than the other two classic algorithms.
基金Construction Program of the Key Discipline of State Administration of Traditional Chinese Medicine of China(ZYYZDXK-2023069)Research Project of Shanghai Municipal Health Commission (2024QN018)Shanghai University of Traditional Chinese Medicine Science and Technology Development Program (23KFL005)。
文摘Objective To develop a non-invasive predictive model for coronary artery stenosis severity based on adaptive multi-modal integration of traditional Chinese and western medicine data.Methods Clinical indicators,echocardiographic data,traditional Chinese medicine(TCM)tongue manifestations,and facial features were collected from patients who underwent coro-nary computed tomography angiography(CTA)in the Cardiac Care Unit(CCU)of Shanghai Tenth People's Hospital between May 1,2023 and May 1,2024.An adaptive weighted multi-modal data fusion(AWMDF)model based on deep learning was constructed to predict the severity of coronary artery stenosis.The model was evaluated using metrics including accura-cy,precision,recall,F1 score,and the area under the receiver operating characteristic(ROC)curve(AUC).Further performance assessment was conducted through comparisons with six ensemble machine learning methods,data ablation,model component ablation,and various decision-level fusion strategies.Results A total of 158 patients were included in the study.The AWMDF model achieved ex-cellent predictive performance(AUC=0.973,accuracy=0.937,precision=0.937,recall=0.929,and F1 score=0.933).Compared with model ablation,data ablation experiments,and various traditional machine learning models,the AWMDF model demonstrated superior per-formance.Moreover,the adaptive weighting strategy outperformed alternative approaches,including simple weighting,averaging,voting,and fixed-weight schemes.Conclusion The AWMDF model demonstrates potential clinical value in the non-invasive prediction of coronary artery disease and could serve as a tool for clinical decision support.
文摘Traditional Chinese medicine(TCM)demonstrates distinctive advantages in disease prevention and treatment.However,analyzing its biological mechanisms through the modern medical research paradigm of“single drug,single target”presents significant challenges due to its holistic approach.Network pharmacology and its core theory of network targets connect drugs and diseases from a holistic and systematic perspective based on biological networks,overcoming the limitations of reductionist research models and showing considerable value in TCM research.Recent integration of network target computational and experimental methods with artificial intelligence(AI)and multi-modal multi-omics technologies has substantially enhanced network pharmacology methodology.The advancement in computational and experimental techniques provides complementary support for network target theory in decoding TCM principles.This review,centered on network targets,examines the progress of network target methods combined with AI in predicting disease molecular mechanisms and drug-target relationships,alongside the application of multi-modal multi-omics technologies in analyzing TCM formulae,syndromes,and toxicity.Looking forward,network target theory is expected to incorporate emerging technologies while developing novel approaches aligned with its unique characteristics,potentially leading to significant breakthroughs in TCM research and advancing scientific understanding and innovation in TCM.
基金supported by the National Natural Science Foundation of China(Grant Nos.62071315 and 62271336).
文摘The multi-modal characteristics of mineral particles play a pivotal role in enhancing the classification accuracy,which is critical for obtaining a profound understanding of the Earth's composition and ensuring effective exploitation utilization of its resources.However,the existing methods for classifying mineral particles do not fully utilize these multi-modal features,thereby limiting the classification accuracy.Furthermore,when conventional multi-modal image classification methods are applied to planepolarized and cross-polarized sequence images of mineral particles,they encounter issues such as information loss,misaligned features,and challenges in spatiotemporal feature extraction.To address these challenges,we propose a multi-modal mineral particle polarization image classification network(MMGC-Net)for precise mineral particle classification.Initially,MMGC-Net employs a two-dimensional(2D)backbone network with shared parameters to extract features from two types of polarized images to ensure feature alignment.Subsequently,a cross-polarized intra-modal feature fusion module is designed to refine the spatiotemporal features from the extracted features of the cross-polarized sequence images.Ultimately,the inter-modal feature fusion module integrates the two types of modal features to enhance the classification precision.Quantitative and qualitative experimental results indicate that when compared with the current state-of-the-art multi-modal image classification methods,MMGC-Net demonstrates marked superiority in terms of mineral particle multi-modal feature learning and four classification evaluation metrics.It also demonstrates better stability than the existing models.
基金supported by the National Natural Science Foundation of China(Nos.62371323,62401380,U2433217,U2333209,and U20A20161)Natural Science Foundation of Sichuan Province,China(Nos.2025ZNSFSC1476)+2 种基金Sichuan Science and Technology Program,China(Nos.2024YFG0010 and 2024ZDZX0046)the Institutional Research Fund from Sichuan University(Nos.2024SCUQJTX030)the Open Fund of Key Laboratory of Flight Techniques and Flight Safety,CAAC(Nos.GY2024-01A).
文摘With the advent of the next-generation Air Traffic Control(ATC)system,there is growing interest in using Artificial Intelligence(AI)techniques to enhance Situation Awareness(SA)for ATC Controllers(ATCOs),i.e.,Intelligent SA(ISA).However,the existing AI-based SA approaches often rely on unimodal data and lack a comprehensive description and benchmark of the ISA tasks utilizing multi-modal data for real-time ATC environments.To address this gap,by analyzing the situation awareness procedure of the ATCOs,the ISA task is refined to the processing of the two primary elements,i.e.,spoken instructions and flight trajectories.Subsequently,the ISA is further formulated into Controlling Intent Understanding(CIU)and Flight Trajectory Prediction(FTP)tasks.For the CIU task,an innovative automatic speech recognition and understanding framework is designed to extract the controlling intent from unstructured and continuous ATC communications.For the FTP task,the single-and multi-horizon FTP approaches are investigated to support the high-precision prediction of the situation evolution.A total of 32 unimodal/multi-modal advanced methods with extensive evaluation metrics are introduced to conduct the benchmarks on the real-world multi-modal ATC situation dataset.Experimental results demonstrate the effectiveness of AI-based techniques in enhancing ISA for the ATC environment.
基金Shanghai Frontier Science Research Center for Modern Textiles,Donghua University,ChinaOpen Project of Henan Key Laboratory of Intelligent Manufacturing of Mechanical Equipment,Zhengzhou University of Light Industry,China(No.IM202303)National Key Research and Development Program of China(No.2019YFB1706300)。
文摘A personalized outfit recommendation has emerged as a hot research topic in the fashion domain.However,existing recommendations do not fully exploit user style preferences.Typically,users prefer particular styles such as casual and athletic styles,and consider attributes like color and texture when selecting outfits.To achieve personalized outfit recommendations in line with user style preferences,this paper proposes a personal style guided outfit recommendation with multi-modal fashion compatibility modeling,termed as PSGNet.Firstly,a style classifier is designed to categorize fashion images of various clothing types and attributes into distinct style categories.Secondly,a personal style prediction module extracts user style preferences by analyzing historical data.Then,to address the limitations of single-modal representations and enhance fashion compatibility,both fashion images and text data are leveraged to extract multi-modal features.Finally,PSGNet integrates these components through Bayesian personalized ranking(BPR)to unify the personal style and fashion compatibility,where the former is used as personal style features and guides the output of the personalized outfit recommendation tailored to the target user.Extensive experiments on large-scale datasets demonstrate that the proposed model is efficient on the personalized outfit recommendation.
文摘With the rapid progress of the artificial intelligence(AI)technology and mobile internet,3D hand pose estimation has become critical to various intelligent application areas,e.g.,human-computer interaction.To avoid the low accuracy of single-modal estimation and the high complexity of traditional multi-modal 3D estimation,this paper proposes a novel multi-modal multi-view(MMV)3D hand pose estimation system,which introduces a registration before translation(RT)-translation before registration(TR)jointed conditional generative adversarial network(cGAN)to train a multi-modal registration network,and then employs the multi-modal feature fusion to achieve high-quality estimation,with low hardware and software costs both in data acquisition and processing.Experimental results demonstrate that the MMV system is effective and feasible in various scenarios.It is promising for the MMV system to be used in broad intelligent application areas.
基金funded by Research Project,grant number BHQ090003000X03.
文摘Multi-modal Named Entity Recognition(MNER)aims to better identify meaningful textual entities by integrating information from images.Previous work has focused on extracting visual semantics at a fine-grained level,or obtaining entity related external knowledge from knowledge bases or Large Language Models(LLMs).However,these approaches ignore the poor semantic correlation between visual and textual modalities in MNER datasets and do not explore different multi-modal fusion approaches.In this paper,we present MMAVK,a multi-modal named entity recognition model with auxiliary visual knowledge and word-level fusion,which aims to leverage the Multi-modal Large Language Model(MLLM)as an implicit knowledge base.It also extracts vision-based auxiliary knowledge from the image formore accurate and effective recognition.Specifically,we propose vision-based auxiliary knowledge generation,which guides the MLLM to extract external knowledge exclusively derived from images to aid entity recognition by designing target-specific prompts,thus avoiding redundant recognition and cognitive confusion caused by the simultaneous processing of image-text pairs.Furthermore,we employ a word-level multi-modal fusion mechanism to fuse the extracted external knowledge with each word-embedding embedded from the transformerbased encoder.Extensive experimental results demonstrate that MMAVK outperforms or equals the state-of-the-art methods on the two classical MNER datasets,even when the largemodels employed have significantly fewer parameters than other baselines.
基金funded by Research Project,grant number BHQ090003000X03。
文摘Multi-modal knowledge graph completion(MMKGC)aims to complete missing entities or relations in multi-modal knowledge graphs,thereby discovering more previously unknown triples.Due to the continuous growth of data and knowledge and the limitations of data sources,the visual knowledge within the knowledge graphs is generally of low quality,and some entities suffer from the issue of missing visual modality.Nevertheless,previous studies of MMKGC have primarily focused on how to facilitate modality interaction and fusion while neglecting the problems of low modality quality and modality missing.In this case,mainstream MMKGC models only use pre-trained visual encoders to extract features and transfer the semantic information to the joint embeddings through modal fusion,which inevitably suffers from problems such as error propagation and increased uncertainty.To address these problems,we propose a Multi-modal knowledge graph Completion model based on Super-resolution and Detailed Description Generation(MMCSD).Specifically,we leverage a pre-trained residual network to enhance the resolution and improve the quality of the visual modality.Moreover,we design multi-level visual semantic extraction and entity description generation,thereby further extracting entity semantics from structural triples and visual images.Meanwhile,we train a variational multi-modal auto-encoder and utilize a pre-trained multi-modal language model to complement the missing visual features.We conducted experiments on FB15K-237 and DB13K,and the results showed that MMCSD can effectively perform MMKGC and achieve state-of-the-art performance.
基金supported by the Deanship of Research and Graduate Studies at King Khalid University under Small Research Project grant number RGP1/139/45.
文摘Integrating multiple medical imaging techniques,including Magnetic Resonance Imaging(MRI),Computed Tomography,Positron Emission Tomography(PET),and ultrasound,provides a comprehensive view of the patient health status.Each of these methods contributes unique diagnostic insights,enhancing the overall assessment of patient condition.Nevertheless,the amalgamation of data from multiple modalities presents difficulties due to disparities in resolution,data collection methods,and noise levels.While traditional models like Convolutional Neural Networks(CNNs)excel in single-modality tasks,they struggle to handle multi-modal complexities,lacking the capacity to model global relationships.This research presents a novel approach for examining multi-modal medical imagery using a transformer-based system.The framework employs self-attention and cross-attention mechanisms to synchronize and integrate features across various modalities.Additionally,it shows resilience to variations in noise and image quality,making it adaptable for real-time clinical use.To address the computational hurdles linked to transformer models,particularly in real-time clinical applications in resource-constrained environments,several optimization techniques have been integrated to boost scalability and efficiency.Initially,a streamlined transformer architecture was adopted to minimize the computational load while maintaining model effectiveness.Methods such as model pruning,quantization,and knowledge distillation have been applied to reduce the parameter count and enhance the inference speed.Furthermore,efficient attention mechanisms such as linear or sparse attention were employed to alleviate the substantial memory and processing requirements of traditional self-attention operations.For further deployment optimization,researchers have implemented hardware-aware acceleration strategies,including the use of TensorRT and ONNX-based model compression,to ensure efficient execution on edge devices.These optimizations allow the approach to function effectively in real-time clinical settings,ensuring viability even in environments with limited resources.Future research directions include integrating non-imaging data to facilitate personalized treatment and enhancing computational efficiency for implementation in resource-limited environments.This study highlights the transformative potential of transformer models in multi-modal medical imaging,offering improvements in diagnostic accuracy and patient care outcomes.
