Potential high-temperature risks exist in heat-prone components of electric moped charging devices,such as sockets,interfaces,and controllers.Traditional detection methods have limitations in terms of real-time perfor...Potential high-temperature risks exist in heat-prone components of electric moped charging devices,such as sockets,interfaces,and controllers.Traditional detection methods have limitations in terms of real-time performance and monitoring scope.To address this,a temperature detection method based on infrared image processing has been proposed:utilizing the median filtering algorithm to denoise the original infrared image,then applying an image segmentation algorithm to divide the image.展开更多
Existing semi-supervisedmedical image segmentation algorithms use copy-paste data augmentation to correct the labeled-unlabeled data distribution mismatch.However,current copy-paste methods have three limitations:(1)t...Existing semi-supervisedmedical image segmentation algorithms use copy-paste data augmentation to correct the labeled-unlabeled data distribution mismatch.However,current copy-paste methods have three limitations:(1)training the model solely with copy-paste mixed pictures from labeled and unlabeled input loses a lot of labeled information;(2)low-quality pseudo-labels can cause confirmation bias in pseudo-supervised learning on unlabeled data;(3)the segmentation performance in low-contrast and local regions is less than optimal.We design a Stochastic Augmentation-Based Dual-Teaching Auxiliary Training Strategy(SADT),which enhances feature diversity and learns high-quality features to overcome these problems.To be more precise,SADT trains the Student Network by using pseudo-label-based training from Teacher Network 1 and supervised learning with labeled data,which prevents the loss of rare labeled data.We introduce a bi-directional copy-pastemask with progressive high-entropy filtering to reduce data distribution disparities and mitigate confirmation bias in pseudo-supervision.For the mixed images,Deep-Shallow Spatial Contrastive Learning(DSSCL)is proposed in the feature spaces of Teacher Network 2 and the Student Network to improve the segmentation capabilities in low-contrast and local areas.In this procedure,the features retrieved by the Student Network are subjected to a random feature perturbation technique.On two openly available datasets,extensive trials show that our proposed SADT performs much better than the state-ofthe-art semi-supervised medical segmentation techniques.Using only 10%of the labeled data for training,SADT was able to acquire a Dice score of 90.10%on the ACDC(Automatic Cardiac Diagnosis Challenge)dataset.展开更多
Deep learning now underpins many state-of-the-art systems for biomedical image and signal processing,enabling automated lesion detection,physiological monitoring,and therapy planning with accuracy that rivals expert p...Deep learning now underpins many state-of-the-art systems for biomedical image and signal processing,enabling automated lesion detection,physiological monitoring,and therapy planning with accuracy that rivals expert performance.This survey reviews the principal model families as convolutional,recurrent,generative,reinforcement,autoencoder,and transfer-learning approaches as emphasising how their architectural choices map to tasks such as segmentation,classification,reconstruction,and anomaly detection.A dedicated treatment of multimodal fusion networks shows how imaging features can be integrated with genomic profiles and clinical records to yield more robust,context-aware predictions.To support clinical adoption,we outline post-hoc explainability techniques(Grad-CAM,SHAP,LIME)and describe emerging intrinsically interpretable designs that expose decision logic to end users.Regulatory guidance from the U.S.FDA,the European Medicines Agency,and the EU AI Act is summarised,linking transparency and lifecycle-monitoring requirements to concrete development practices.Remaining challenges as data imbalance,computational cost,privacy constraints,and cross-domain generalization are discussed alongside promising solutions such as federated learning,uncertainty quantification,and lightweight 3-D architectures.The article therefore offers researchers,clinicians,and policymakers a concise,practice-oriented roadmap for deploying trustworthy deep-learning systems in healthcare.展开更多
Medical image segmentation has become a cornerstone for many healthcare applications,allowing for the automated extraction of critical information from images such as Computed Tomography(CT)scans,Magnetic Resonance Im...Medical image segmentation has become a cornerstone for many healthcare applications,allowing for the automated extraction of critical information from images such as Computed Tomography(CT)scans,Magnetic Resonance Imaging(MRIs),and X-rays.The introduction of U-Net in 2015 has significantly advanced segmentation capabilities,especially for small datasets commonly found in medical imaging.Since then,various modifications to the original U-Net architecture have been proposed to enhance segmentation accuracy and tackle challenges like class imbalance,data scarcity,and multi-modal image processing.This paper provides a detailed review and comparison of several U-Net-based architectures,focusing on their effectiveness in medical image segmentation tasks.We evaluate performance metrics such as Dice Similarity Coefficient(DSC)and Intersection over Union(IoU)across different U-Net variants including HmsU-Net,CrossU-Net,mResU-Net,and others.Our results indicate that architectural enhancements such as transformers,attention mechanisms,and residual connections improve segmentation performance across diverse medical imaging applications,including tumor detection,organ segmentation,and lesion identification.The study also identifies current challenges in the field,including data variability,limited dataset sizes,and issues with class imbalance.Based on these findings,the paper suggests potential future directions for improving the robustness and clinical applicability of U-Net-based models in medical image segmentation.展开更多
Image segmentation is attracting increasing attention in the field of medical image analysis.Since widespread utilization across various medical applications,ensuring and improving segmentation accuracy has become a c...Image segmentation is attracting increasing attention in the field of medical image analysis.Since widespread utilization across various medical applications,ensuring and improving segmentation accuracy has become a crucial topic of research.With advances in deep learning,researchers have developed numerous methods that combine Transformers and convolutional neural networks(CNNs)to create highly accurate models for medical image segmentation.However,efforts to further enhance accuracy by developing larger and more complex models or training with more extensive datasets,significantly increase computational resource consumption.To address this problem,we propose BiCLIP-nnFormer(the prefix"Bi"refers to the use of two distinct CLIP models),a virtual multimodal instrument that leverages CLIP models to enhance the segmentation performance of a medical segmentation model nnFormer.Since two CLIP models(PMC-CLIP and CoCa-CLIP)are pre-trained on large datasets,they do not require additional training,thus conserving computation resources.These models are used offline to extract image and text embeddings from medical images.These embeddings are then processed by the proposed 3D CLIP adapter,which adapts the CLIP knowledge for segmentation tasks by fine-tuning.Finally,the adapted embeddings are fused with feature maps extracted from the nnFormer encoder for generating predicted masks.This process enriches the representation capabilities of the feature maps by integrating global multimodal information,leading to more precise segmentation predictions.We demonstrate the superiority of BiCLIP-nnFormer and the effectiveness of using CLIP models to enhance nnFormer through experiments on two public datasets,namely the Synapse multi-organ segmentation dataset(Synapse)and the Automatic Cardiac Diagnosis Challenge dataset(ACDC),as well as a self-annotated lung multi-category segmentation dataset(LMCS).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
Agromyzid leafminers cause significant economic losses in both vegetable and horticultural crops,and precise assessments of pesticide needs must be based on the extent of leaf damage.Traditionally,surveyors estimate t...Agromyzid leafminers cause significant economic losses in both vegetable and horticultural crops,and precise assessments of pesticide needs must be based on the extent of leaf damage.Traditionally,surveyors estimate the damage by visually comparing the proportion of damaged to intact leaf area,a method that lacks objectivity,precision,and reliable data traceability.To address these issues,an advanced survey system that combines augmented reality(AR)glasses with a camera and an artificial intelligence(AI)algorithm was developed in this study to objectively and accurately assess leafminer damage in the feld.By wearing AR glasses equipped with a voice-controlled camera,surveyors can easily flatten damaged leaves by hand and capture images for analysis.This method can provide a precise and reliable diagnosis of leafminer damage levels,which in turn supports the implementation of scientifically grounded and targeted pest management strategies.To calculate the leafminer damage level,the DeepLab-Leafminer model was proposed to precisely segment the leafminer-damaged regions and the intact leaf region.The integration of an edge-aware module and a Canny loss function into the DeepLabv3+model enhanced the DeepLab-Leafminer model's capability to accurately segment the edges of leafminer-damaged regions,which often exhibit irregular shapes.Compared with state-of-the-art segmentation models,the DeepLabLeafminer model achieved superior segmentation performance with an Intersection over Union(IoU)of 81.23%and an F1score of 87.92%on leafminer-damaged leaves.The test results revealed a 92.38%diagnosis accuracy of leafminer damage levels based on the DeepLab-Leafminer model.A mobile application and a web platform were developed to assist surveyors in displaying the diagnostic results of leafminer damage levels.This system provides surveyors with an advanced,user-friendly,and accurate tool for assessing agromyzid leafminer damage in agricultural felds using wearable AR glasses and an AI model.This method can also be utilized to automatically diagnose pest and disease damage levels in other crops based on leaf images.展开更多
The traditional EnFCM(Enhanced fuzzy C-means)algorithm only considers the grey-scale features in image segmentation,resulting in less than satisfactory results when the algorithm is used for remote sensing woodland im...The traditional EnFCM(Enhanced fuzzy C-means)algorithm only considers the grey-scale features in image segmentation,resulting in less than satisfactory results when the algorithm is used for remote sensing woodland image segmentation and extraction.An EnFCM remote sensing forest land extraction method based on PCA multi-feature fusion was proposed.Firstly,histogram equalization was applied to improve the image contrast.Secondly,the texture and edge features of the image were extracted,and a multi-feature fused pixel image was generated using the PCA technique.Moreover,the fused feature was used as a feature constraint to measure the difference of pixels instead of a single grey-scale feature.Finally,an improved feature distance metric calculated the similarity between the pixel points and the cluster center to complete the cluster segmentation.The experimental results showed that the error was between 1.5%and 4.0%compared with the forested area counted by experts’hand-drawing,which could obtain a high accuracy segmentation and extraction result.展开更多
The self-attention mechanism of Transformers,which captures long-range contextual information,has demonstrated significant potential in image segmentation.However,their ability to learn local,contextual relationships ...The self-attention mechanism of Transformers,which captures long-range contextual information,has demonstrated significant potential in image segmentation.However,their ability to learn local,contextual relationships between pixels requires further improvement.Previous methods face challenges in efficiently managing multi-scale fea-tures of different granularities from the encoder backbone,leaving room for improvement in their global representation and feature extraction capabilities.To address these challenges,we propose a novel Decoder with Multi-Head Feature Receptors(DMHFR),which receives multi-scale features from the encoder backbone and organizes them into three feature groups with different granularities:coarse,fine-grained,and full set.These groups are subsequently processed by Multi-Head Feature Receptors(MHFRs)after feature capture and modeling operations.MHFRs include two Three-Head Feature Receptors(THFRs)and one Four-Head Feature Receptor(FHFR).Each group of features is passed through these MHFRs and then fed into axial transformers,which help the model capture long-range dependencies within the features.The three MHFRs produce three distinct feature outputs.The output from the FHFR serves as auxiliary auxiliary features in the prediction head,and the prediction output and their losses will eventually be aggregated.Experimental results show that the Transformer using DMHFR outperforms 15 state of the arts(SOTA)methods on five public datasets.Specifically,it achieved significant improvements in mean DICE scores over the classic Parallel Reverse Attention Network(PraNet)method,with gains of 4.1%,2.2%,1.4%,8.9%,and 16.3%on the CVC-ClinicDB,Kvasir-SEG,CVC-T,CVC-ColonDB,and ETIS-LaribPolypDB datasets,respectively.展开更多
Separation between water and land is vital for marine scientific research and coastal zone planning and management.The contrasting backscatter properties of land and ocean enable clear water edge line identification i...Separation between water and land is vital for marine scientific research and coastal zone planning and management.The contrasting backscatter properties of land and ocean enable clear water edge line identification in synthetic aperture radar(SAR)imagery.However,SAR images are prone to speckle noise,and the complexity of the water-land boundaries environment makes accurate water-land separation challenging.To overcome noise and complex background interference in remote sensing images,an improved level set method was employed to enhance water-land separation.In the traditional distance regularized level set method,the selection of the image correlation weight coefficient and the edge indicator function directly influences the accuracy of the final segmentation results.A novel level set segmentation algorithm incorporating an improved edge indicator function is proposed to efficiently and accurately separate the water edge lines in SAR images.The distance regularized level set evolution model is enhanced by incorporating the signed pressure force function as an adaptive parameter,which serves as an external constraint for curve evolution.A novel level set model with an adaptive edge indicator function,combining gradient and regional information,is proposed.Experimental results demonstrate that the proposed model enhances the accuracy of waterland separation in SAR images.However,further research is needed to evaluate its potential for detecting boundaries in diverse marine environments and across different types of SAR imagery.展开更多
In the field of medical image processing,combining global and local relationship modeling constitutes an effective strategy for precise segmentation.Prior research has established the validity of Convolutional Neural ...In the field of medical image processing,combining global and local relationship modeling constitutes an effective strategy for precise segmentation.Prior research has established the validity of Convolutional Neural Networks(CNN)in modeling local relationships.Conversely,Transformers have demonstrated their capability to effectively capture global contextual information.However,when utilized to address CNNs’limitations in modeling global relationships,Transformers are hindered by substantial computational complexity.To address this issue,we introduce Mamba,a State-Space Model(SSM)that exhibits exceptional proficiency in modeling long-range dependencies in sequential data.Given Mamba’s demonstrated potential in 2D medical image segmentation in previous studies,we have designed a Dual-encoder Global-local Feature Extraction Network based on Mamba,termed DGFE-Mamba,to accurately capture and fuse long-range dependencies and local dependencies within multi-scale features.Compared to Transformer-based methods,the DGFE-Mamba model excels in comprehensive feature modeling and demonstrates significantly improved segmentation accuracy.To validate the effectiveness and practicality of DGFE-Mamba,we conducted tests on the Automatic Cardiac Diagnosis Challenge(ACDC)dataset,the Synapse multi-organ CT abdominal segmentation dataset,and the Colorectal Cancer Clinic(CVC-ClinicDB)dataset.The results showed that DGFE-Mamba achieved Dice coefficients of 92.20,83.67,and 94.13,respectively.These findings comprehensively validate the effectiveness and practicality of the proposed DGFE-Mamba architecture.展开更多
Various and intricate varieties of lung disease have made it challenging for computer aided diagnosis to appropriately segment lung lesions utilizing computed tomography(CT)images.This study integrates transfer learni...Various and intricate varieties of lung disease have made it challenging for computer aided diagnosis to appropriately segment lung lesions utilizing computed tomography(CT)images.This study integrates transfer learning with the attention mechanism to construct a deep learning model that can automatically detect new coronary pneumonia on lung CT images.In this study,using VGG16 pre-trained by ImageNet as the encoder,the decoder was established utilizing the U-Net structure.The attention module is incorporated during each concatenate procedure,permitting the model to concentrate on the critical information and identify the crucial components efficiently.The public COVID-19-CT-Seg-Benchmark dataset was utilized for experiments,and the highest scores for Dice,F1,and Accuracy were 0.9071,0.9076,and 0.9965,respectively.The generalization performance was assessed concurrently,with performance metrics including Dice,F1,and Accuracy over 0.8.The experimental findings indicate the feasibility of the segmentation network proposed in this study.展开更多
A growing number of skin laser treatments have rapidly evolved and increased their role in the field of dermatology,laser treatment is considered to be used for a variety of pigmentary dermatosis as well as aesthetic ...A growing number of skin laser treatments have rapidly evolved and increased their role in the field of dermatology,laser treatment is considered to be used for a variety of pigmentary dermatosis as well as aesthetic problems.The standardized assessment of laser treatment efficacy is crucial for the interpretation and comparison of studies related to laser treatment of skin disorders.In this study,we propose an evaluation method to quantitatively assess laser treatment efficacy based on the image segmentation technology.A tattoo model of Sprague Dawley(SD)rats was established and treated by picosecond laser treatments at varying energy levels.Images of the tattoo models were captured before and after laser treatment,and feature extraction was conducted to quantify the tattooed area and pigment gradation.Subsequently,the clearance rate,which has been a standardized parameter,was calculated.The results indicate that the clearance rates obtained through this quantitative algorithm are comparable and exhibit smaller standard deviations compared with scale scores(4.59%versus 7.93%in the low-energy group,4.01%versus 9.05%in the medium-energy group,and 4.29%versus 10.23%in the high-energy group).This underscores the greater accuracy,objectivity,and reproducibility in assessing treatment responses.The quantitative evaluation of pigment removal holds promise for facilitating faster and more robust assessments in research and development.Additionally,it may enable the optimization of treatments tailored to individual patients,thereby contributing to more effective and personalized dermatological care.展开更多
Despite its remarkable performance on natural images,the segment anything model(SAM)lacks domain-specific information in medical imaging.and faces the challenge of losing local multi-scale information in the encoding ...Despite its remarkable performance on natural images,the segment anything model(SAM)lacks domain-specific information in medical imaging.and faces the challenge of losing local multi-scale information in the encoding phase.This paper presents a medical image segmentation model based on SAM with a local multi-scale feature encoder(LMSFE-SAM)to address the issues above.Firstly,based on the SAM,a local multi-scale feature encoder is introduced to improve the representation of features within local receptive field,thereby supplying the Vision Transformer(ViT)branch in SAM with enriched local multi-scale contextual information.At the same time,a multiaxial Hadamard product module(MHPM)is incorporated into the local multi-scale feature encoder in a lightweight manner to reduce the quadratic complexity and noise interference.Subsequently,a cross-branch balancing adapter is designed to balance the local and global information between the local multi-scale feature encoder and the ViT encoder in SAM.Finally,to obtain smaller input image size and to mitigate overlapping in patch embeddings,the size of the input image is reduced from 1024×1024 pixels to 256×256 pixels,and a multidimensional information adaptation component is developed,which includes feature adapters,position adapters,and channel-spatial adapters.This component effectively integrates the information from small-sized medical images into SAM,enhancing its suitability for clinical deployment.The proposed model demonstrates an average enhancement ranging from 0.0387 to 0.3191 across six objective evaluation metrics on BUSI,DDTI,and TN3K datasets compared to eight other representative image segmentation models.This significantly enhances the performance of the SAM on medical images,providing clinicians with a powerful tool in clinical diagnosis.展开更多
Foundation models(FMs)have rapidly evolved and have achieved signicant accomplishments in computer vision tasks.Specically,the prompt mechanism conveniently allows users to integrate image prior information into the m...Foundation models(FMs)have rapidly evolved and have achieved signicant accomplishments in computer vision tasks.Specically,the prompt mechanism conveniently allows users to integrate image prior information into the model,making it possible to apply models without any training.Therefore,we proposed a workflow based on foundation models and zero training to solve the tasks of photoacoustic(PA)image processing.We employed the Segment Anything Model(SAM)by setting simple prompts and integrating the model's outputs with prior knowledge of the imaged objects to accomplish various tasks,including:(1)removing the skin signal in three-dimensional PA image rendering;(2)dual speed-of-sound reconstruction,and(3)segmentation ofnger blood vessels.Through these demonstrations,we have concluded that FMs can be directly applied in PA imaging without the requirement for network design and training.This potentially allows for a hands-on,convenient approach to achieving efficient and accurate segmentation of PA images.This paper serves as a comprehensive tutorial,facilitating the mastery of the technique through the provision of code and sample datasets.展开更多
The visual noise of each light intensity area is different when the image is drawn by Monte Carlo method.However,the existing denoising algorithms have limited denoising performance under complex lighting conditions a...The visual noise of each light intensity area is different when the image is drawn by Monte Carlo method.However,the existing denoising algorithms have limited denoising performance under complex lighting conditions and are easy to lose detailed information.So we propose a rendered image denoising method with filtering guided by lighting information.First,we design an image segmentation algorithm based on lighting information to segment the image into different illumination areas.Then,we establish the parameter prediction model guided by lighting information for filtering(PGLF)to predict the filtering parameters of different illumination areas.For different illumination areas,we use these filtering parameters to construct area filters,and the filters are guided by the lighting information to perform sub-area filtering.Finally,the filtering results are fused with auxiliary features to output denoised images for improving the overall denoising effect of the image.Under the physically based rendering tool(PBRT)scene and Tungsten dataset,the experimental results show that compared with other guided filtering denoising methods,our method improves the peak signal-to-noise ratio(PSNR)metrics by 4.2164 dB on average and the structural similarity index(SSIM)metrics by 7.8%on average.This shows that our method can better reduce the noise in complex lighting scenesand improvethe imagequality.展开更多
Data augmentation plays an important role in training deep neural model by expanding the size and diversity of the dataset.Initially,data augmentation mainly involved some simple transformations of images.Later,in ord...Data augmentation plays an important role in training deep neural model by expanding the size and diversity of the dataset.Initially,data augmentation mainly involved some simple transformations of images.Later,in order to increase the diversity and complexity of data,more advanced methods appeared and evolved to sophisticated generative models.However,these methods required a mass of computation of training or searching.In this paper,a novel training-free method that utilises the Pre-Trained Segment Anything Model(SAM)model as a data augmentation tool(PTSAM-DA)is proposed to generate the augmented annotations for images.Without the need for training,it obtains prompt boxes from the original annotations and then feeds the boxes to the pre-trained SAM to generate diverse and improved annotations.In this way,annotations are augmented more ingenious than simple manipulations without incurring huge computation for training a data augmentation model.Multiple comparative experiments on three datasets are conducted,including an in-house dataset,ADE20K and COCO2017.On this in-house dataset,namely Agricultural Plot Segmentation Dataset,maximum improvements of 3.77%and 8.92%are gained in two mainstream metrics,mIoU and mAcc,respectively.Consequently,large vision models like SAM are proven to be promising not only in image segmentation but also in data augmentation.展开更多
Deep learning(DL),derived from the domain of Artificial Neural Networks(ANN),forms one of the most essential components of modern deep learning algorithms.DL segmentation models rely on layer-by-layer convolution-base...Deep learning(DL),derived from the domain of Artificial Neural Networks(ANN),forms one of the most essential components of modern deep learning algorithms.DL segmentation models rely on layer-by-layer convolution-based feature representation,guided by forward and backward propagation.Acritical aspect of this process is the selection of an appropriate activation function(AF)to ensure robustmodel learning.However,existing activation functions often fail to effectively address the vanishing gradient problem or are complicated by the need for manual parameter tuning.Most current research on activation function design focuses on classification tasks using natural image datasets such asMNIST,CIFAR-10,and CIFAR-100.To address this gap,this study proposesMed-ReLU,a novel activation function specifically designed for medical image segmentation.Med-ReLU prevents deep learning models fromsuffering dead neurons or vanishing gradient issues.It is a hybrid activation function that combines the properties of ReLU and Softsign.For positive inputs,Med-ReLU adopts the linear behavior of ReLU to avoid vanishing gradients,while for negative inputs,it exhibits the Softsign’s polynomial convergence,ensuring robust training and avoiding inactive neurons across the training set.The training performance and segmentation accuracy ofMed-ReLU have been thoroughly evaluated,demonstrating stable learning behavior and resistance to overfitting.It consistently outperforms state-of-the-art activation functions inmedical image segmentation tasks.Designed as a parameter-free function,Med-ReLU is simple to implement in complex deep learning architectures,and its effectiveness spans various neural network models and anomaly detection scenarios.展开更多
In the realm of medical image segmentation,particularly in cardiac magnetic resonance imaging(MRI),achieving robust performance with limited annotated data is a significant challenge.Performance often degrades when fa...In the realm of medical image segmentation,particularly in cardiac magnetic resonance imaging(MRI),achieving robust performance with limited annotated data is a significant challenge.Performance often degrades when faced with testing scenarios from unknown domains.To address this problem,this paper proposes a novel semi-supervised approach for cardiac magnetic resonance image segmentation,aiming to enhance predictive capabilities and domain generalization(DG).This paper establishes an MT-like model utilizing pseudo-labeling and consistency regularization from semi-supervised learning,and integrates uncertainty estimation to improve the accuracy of pseudo-labels.Additionally,to tackle the challenge of domain generalization,a data manipulation strategy is introduced,extracting spatial and content-related information from images across different domains,enriching the dataset with a multi-domain perspective.This papers method is meticulously evaluated on the publicly available cardiac magnetic resonance imaging dataset M&Ms,validating its effectiveness.Comparative analyses against various methods highlight the out-standing performance of this papers approach,demonstrating its capability to segment cardiac magnetic resonance images in previously unseen domains even with limited annotated data.展开更多
Medical image segmentation,i.e.,labeling structures of interest in medical images,is crucial for disease diagnosis and treatment in radiology.In reversible data hiding in medical images(RDHMI),segmentation consists of...Medical image segmentation,i.e.,labeling structures of interest in medical images,is crucial for disease diagnosis and treatment in radiology.In reversible data hiding in medical images(RDHMI),segmentation consists of only two regions:the focal and nonfocal regions.The focal region mainly contains information for diagnosis,while the nonfocal region serves as the monochrome background.The current traditional segmentation methods utilized in RDHMI are inaccurate for complex medical images,and manual segmentation is time-consuming,poorly reproducible,and operator-dependent.Implementing state-of-the-art deep learning(DL)models will facilitate key benefits,but the lack of domain-specific labels for existing medical datasets makes it impossible.To address this problem,this study provides labels of existing medical datasets based on a hybrid segmentation approach to facilitate the implementation of DL segmentation models in this domain.First,an initial segmentation based on a 33 kernel is performed to analyze×identified contour pixels before classifying pixels into focal and nonfocal regions.Then,several human expert raters evaluate and classify the generated labels into accurate and inaccurate labels.The inaccurate labels undergo manual segmentation by medical practitioners and are scored based on a hierarchical voting scheme before being assigned to the proposed dataset.To ensure reliability and integrity in the proposed dataset,we evaluate the accurate automated labels with manually segmented labels by medical practitioners using five assessment metrics:dice coefficient,Jaccard index,precision,recall,and accuracy.The experimental results show labels in the proposed dataset are consistent with the subjective judgment of human experts,with an average accuracy score of 94%and dice coefficient scores between 90%-99%.The study further proposes a ResNet-UNet with concatenated spatial and channel squeeze and excitation(scSE)architecture for semantic segmentation to validate and illustrate the usefulness of the proposed dataset.The results demonstrate the superior performance of the proposed architecture in accurately separating the focal and nonfocal regions compared to state-of-the-art architectures.Dataset information is released under the following URL:https://www.kaggle.com/lordamoah/datasets(accessed on 31 March 2025).展开更多
基金supported by the National Key Research and Development Project of China(No.2023YFB3709605)the National Natural Science Foundation of China(No.62073193)the National College Student Innovation Training Program(No.202310422122)。
文摘Potential high-temperature risks exist in heat-prone components of electric moped charging devices,such as sockets,interfaces,and controllers.Traditional detection methods have limitations in terms of real-time performance and monitoring scope.To address this,a temperature detection method based on infrared image processing has been proposed:utilizing the median filtering algorithm to denoise the original infrared image,then applying an image segmentation algorithm to divide the image.
基金supported by the Natural Science Foundation of China(No.41804112,author:Chengyun Song).
文摘Existing semi-supervisedmedical image segmentation algorithms use copy-paste data augmentation to correct the labeled-unlabeled data distribution mismatch.However,current copy-paste methods have three limitations:(1)training the model solely with copy-paste mixed pictures from labeled and unlabeled input loses a lot of labeled information;(2)low-quality pseudo-labels can cause confirmation bias in pseudo-supervised learning on unlabeled data;(3)the segmentation performance in low-contrast and local regions is less than optimal.We design a Stochastic Augmentation-Based Dual-Teaching Auxiliary Training Strategy(SADT),which enhances feature diversity and learns high-quality features to overcome these problems.To be more precise,SADT trains the Student Network by using pseudo-label-based training from Teacher Network 1 and supervised learning with labeled data,which prevents the loss of rare labeled data.We introduce a bi-directional copy-pastemask with progressive high-entropy filtering to reduce data distribution disparities and mitigate confirmation bias in pseudo-supervision.For the mixed images,Deep-Shallow Spatial Contrastive Learning(DSSCL)is proposed in the feature spaces of Teacher Network 2 and the Student Network to improve the segmentation capabilities in low-contrast and local areas.In this procedure,the features retrieved by the Student Network are subjected to a random feature perturbation technique.On two openly available datasets,extensive trials show that our proposed SADT performs much better than the state-ofthe-art semi-supervised medical segmentation techniques.Using only 10%of the labeled data for training,SADT was able to acquire a Dice score of 90.10%on the ACDC(Automatic Cardiac Diagnosis Challenge)dataset.
基金supported by the Science Committee of the Ministry of Higher Education and Science of the Republic of Kazakhstan within the framework of grant AP23489899“Applying Deep Learning and Neuroimaging Methods for Brain Stroke Diagnosis”.
文摘Deep learning now underpins many state-of-the-art systems for biomedical image and signal processing,enabling automated lesion detection,physiological monitoring,and therapy planning with accuracy that rivals expert performance.This survey reviews the principal model families as convolutional,recurrent,generative,reinforcement,autoencoder,and transfer-learning approaches as emphasising how their architectural choices map to tasks such as segmentation,classification,reconstruction,and anomaly detection.A dedicated treatment of multimodal fusion networks shows how imaging features can be integrated with genomic profiles and clinical records to yield more robust,context-aware predictions.To support clinical adoption,we outline post-hoc explainability techniques(Grad-CAM,SHAP,LIME)and describe emerging intrinsically interpretable designs that expose decision logic to end users.Regulatory guidance from the U.S.FDA,the European Medicines Agency,and the EU AI Act is summarised,linking transparency and lifecycle-monitoring requirements to concrete development practices.Remaining challenges as data imbalance,computational cost,privacy constraints,and cross-domain generalization are discussed alongside promising solutions such as federated learning,uncertainty quantification,and lightweight 3-D architectures.The article therefore offers researchers,clinicians,and policymakers a concise,practice-oriented roadmap for deploying trustworthy deep-learning systems in healthcare.
文摘Medical image segmentation has become a cornerstone for many healthcare applications,allowing for the automated extraction of critical information from images such as Computed Tomography(CT)scans,Magnetic Resonance Imaging(MRIs),and X-rays.The introduction of U-Net in 2015 has significantly advanced segmentation capabilities,especially for small datasets commonly found in medical imaging.Since then,various modifications to the original U-Net architecture have been proposed to enhance segmentation accuracy and tackle challenges like class imbalance,data scarcity,and multi-modal image processing.This paper provides a detailed review and comparison of several U-Net-based architectures,focusing on their effectiveness in medical image segmentation tasks.We evaluate performance metrics such as Dice Similarity Coefficient(DSC)and Intersection over Union(IoU)across different U-Net variants including HmsU-Net,CrossU-Net,mResU-Net,and others.Our results indicate that architectural enhancements such as transformers,attention mechanisms,and residual connections improve segmentation performance across diverse medical imaging applications,including tumor detection,organ segmentation,and lesion identification.The study also identifies current challenges in the field,including data variability,limited dataset sizes,and issues with class imbalance.Based on these findings,the paper suggests potential future directions for improving the robustness and clinical applicability of U-Net-based models in medical image segmentation.
基金funded by the National Natural Science Foundation of China(Grant No.6240072655)the Hubei Provincial Key Research and Development Program(Grant No.2023BCB151)+1 种基金the Wuhan Natural Science Foundation Exploration Program(Chenguang Program,Grant No.2024040801020202)the Natural Science Foundation of Hubei Province of China(Grant No.2025AFB148).
文摘Image segmentation is attracting increasing attention in the field of medical image analysis.Since widespread utilization across various medical applications,ensuring and improving segmentation accuracy has become a crucial topic of research.With advances in deep learning,researchers have developed numerous methods that combine Transformers and convolutional neural networks(CNNs)to create highly accurate models for medical image segmentation.However,efforts to further enhance accuracy by developing larger and more complex models or training with more extensive datasets,significantly increase computational resource consumption.To address this problem,we propose BiCLIP-nnFormer(the prefix"Bi"refers to the use of two distinct CLIP models),a virtual multimodal instrument that leverages CLIP models to enhance the segmentation performance of a medical segmentation model nnFormer.Since two CLIP models(PMC-CLIP and CoCa-CLIP)are pre-trained on large datasets,they do not require additional training,thus conserving computation resources.These models are used offline to extract image and text embeddings from medical images.These embeddings are then processed by the proposed 3D CLIP adapter,which adapts the CLIP knowledge for segmentation tasks by fine-tuning.Finally,the adapted embeddings are fused with feature maps extracted from the nnFormer encoder for generating predicted masks.This process enriches the representation capabilities of the feature maps by integrating global multimodal information,leading to more precise segmentation predictions.We demonstrate the superiority of BiCLIP-nnFormer and the effectiveness of using CLIP models to enhance nnFormer through experiments on two public datasets,namely the Synapse multi-organ segmentation dataset(Synapse)and the Automatic Cardiac Diagnosis Challenge dataset(ACDC),as well as a self-annotated lung multi-category segmentation dataset(LMCS).
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.
基金supported by the National Key R&D Program of China(2021YFC2600400 and 2023YFC2605200)the National Key Research Program of China(2021YFD1401100)the“San Nong Jiu Fang”Sciences and Technologies Cooperation Project of Zhejiang Province,China(2024SNJF010)。
文摘Agromyzid leafminers cause significant economic losses in both vegetable and horticultural crops,and precise assessments of pesticide needs must be based on the extent of leaf damage.Traditionally,surveyors estimate the damage by visually comparing the proportion of damaged to intact leaf area,a method that lacks objectivity,precision,and reliable data traceability.To address these issues,an advanced survey system that combines augmented reality(AR)glasses with a camera and an artificial intelligence(AI)algorithm was developed in this study to objectively and accurately assess leafminer damage in the feld.By wearing AR glasses equipped with a voice-controlled camera,surveyors can easily flatten damaged leaves by hand and capture images for analysis.This method can provide a precise and reliable diagnosis of leafminer damage levels,which in turn supports the implementation of scientifically grounded and targeted pest management strategies.To calculate the leafminer damage level,the DeepLab-Leafminer model was proposed to precisely segment the leafminer-damaged regions and the intact leaf region.The integration of an edge-aware module and a Canny loss function into the DeepLabv3+model enhanced the DeepLab-Leafminer model's capability to accurately segment the edges of leafminer-damaged regions,which often exhibit irregular shapes.Compared with state-of-the-art segmentation models,the DeepLabLeafminer model achieved superior segmentation performance with an Intersection over Union(IoU)of 81.23%and an F1score of 87.92%on leafminer-damaged leaves.The test results revealed a 92.38%diagnosis accuracy of leafminer damage levels based on the DeepLab-Leafminer model.A mobile application and a web platform were developed to assist surveyors in displaying the diagnostic results of leafminer damage levels.This system provides surveyors with an advanced,user-friendly,and accurate tool for assessing agromyzid leafminer damage in agricultural felds using wearable AR glasses and an AI model.This method can also be utilized to automatically diagnose pest and disease damage levels in other crops based on leaf images.
基金supported by National Natural Science Foundation of China(No.61761027)Gansu Young Doctor’s Fund for Higher Education Institutions(No.2021QB-053)。
文摘The traditional EnFCM(Enhanced fuzzy C-means)algorithm only considers the grey-scale features in image segmentation,resulting in less than satisfactory results when the algorithm is used for remote sensing woodland image segmentation and extraction.An EnFCM remote sensing forest land extraction method based on PCA multi-feature fusion was proposed.Firstly,histogram equalization was applied to improve the image contrast.Secondly,the texture and edge features of the image were extracted,and a multi-feature fused pixel image was generated using the PCA technique.Moreover,the fused feature was used as a feature constraint to measure the difference of pixels instead of a single grey-scale feature.Finally,an improved feature distance metric calculated the similarity between the pixel points and the cluster center to complete the cluster segmentation.The experimental results showed that the error was between 1.5%and 4.0%compared with the forested area counted by experts’hand-drawing,which could obtain a high accuracy segmentation and extraction result.
基金supported by Xiamen Medical and Health Guidance Project in 2021(No.3502Z20214ZD1070)supported by a grant from Guangxi Key Laboratory of Machine Vision and Intelligent Control,China(No.2023B02).
文摘The self-attention mechanism of Transformers,which captures long-range contextual information,has demonstrated significant potential in image segmentation.However,their ability to learn local,contextual relationships between pixels requires further improvement.Previous methods face challenges in efficiently managing multi-scale fea-tures of different granularities from the encoder backbone,leaving room for improvement in their global representation and feature extraction capabilities.To address these challenges,we propose a novel Decoder with Multi-Head Feature Receptors(DMHFR),which receives multi-scale features from the encoder backbone and organizes them into three feature groups with different granularities:coarse,fine-grained,and full set.These groups are subsequently processed by Multi-Head Feature Receptors(MHFRs)after feature capture and modeling operations.MHFRs include two Three-Head Feature Receptors(THFRs)and one Four-Head Feature Receptor(FHFR).Each group of features is passed through these MHFRs and then fed into axial transformers,which help the model capture long-range dependencies within the features.The three MHFRs produce three distinct feature outputs.The output from the FHFR serves as auxiliary auxiliary features in the prediction head,and the prediction output and their losses will eventually be aggregated.Experimental results show that the Transformer using DMHFR outperforms 15 state of the arts(SOTA)methods on five public datasets.Specifically,it achieved significant improvements in mean DICE scores over the classic Parallel Reverse Attention Network(PraNet)method,with gains of 4.1%,2.2%,1.4%,8.9%,and 16.3%on the CVC-ClinicDB,Kvasir-SEG,CVC-T,CVC-ColonDB,and ETIS-LaribPolypDB datasets,respectively.
基金The National Natural Science Foundation of China under contract Nos 61701416 and 61901195Natural Science Foundation of Jiangsu Province of China under contract No.BK20211341China Postdoctoral Science Foundation under contract No.2022M712687.
文摘Separation between water and land is vital for marine scientific research and coastal zone planning and management.The contrasting backscatter properties of land and ocean enable clear water edge line identification in synthetic aperture radar(SAR)imagery.However,SAR images are prone to speckle noise,and the complexity of the water-land boundaries environment makes accurate water-land separation challenging.To overcome noise and complex background interference in remote sensing images,an improved level set method was employed to enhance water-land separation.In the traditional distance regularized level set method,the selection of the image correlation weight coefficient and the edge indicator function directly influences the accuracy of the final segmentation results.A novel level set segmentation algorithm incorporating an improved edge indicator function is proposed to efficiently and accurately separate the water edge lines in SAR images.The distance regularized level set evolution model is enhanced by incorporating the signed pressure force function as an adaptive parameter,which serves as an external constraint for curve evolution.A novel level set model with an adaptive edge indicator function,combining gradient and regional information,is proposed.Experimental results demonstrate that the proposed model enhances the accuracy of waterland separation in SAR images.However,further research is needed to evaluate its potential for detecting boundaries in diverse marine environments and across different types of SAR imagery.
基金supported by the National Natural Science Foundation of China(62276092,62303167)the Postdoctoral Fellowship Program(Grade C)of China Postdoctoral Science Foundation(GZC20230707)+3 种基金the Key Science and Technology Program of Henan Province,China(242102211051)MRC(MC_PC_17171)Royal Society(RP202G0230)BHF(AA/18/3/34220)。
文摘In the field of medical image processing,combining global and local relationship modeling constitutes an effective strategy for precise segmentation.Prior research has established the validity of Convolutional Neural Networks(CNN)in modeling local relationships.Conversely,Transformers have demonstrated their capability to effectively capture global contextual information.However,when utilized to address CNNs’limitations in modeling global relationships,Transformers are hindered by substantial computational complexity.To address this issue,we introduce Mamba,a State-Space Model(SSM)that exhibits exceptional proficiency in modeling long-range dependencies in sequential data.Given Mamba’s demonstrated potential in 2D medical image segmentation in previous studies,we have designed a Dual-encoder Global-local Feature Extraction Network based on Mamba,termed DGFE-Mamba,to accurately capture and fuse long-range dependencies and local dependencies within multi-scale features.Compared to Transformer-based methods,the DGFE-Mamba model excels in comprehensive feature modeling and demonstrates significantly improved segmentation accuracy.To validate the effectiveness and practicality of DGFE-Mamba,we conducted tests on the Automatic Cardiac Diagnosis Challenge(ACDC)dataset,the Synapse multi-organ CT abdominal segmentation dataset,and the Colorectal Cancer Clinic(CVC-ClinicDB)dataset.The results showed that DGFE-Mamba achieved Dice coefficients of 92.20,83.67,and 94.13,respectively.These findings comprehensively validate the effectiveness and practicality of the proposed DGFE-Mamba architecture.
基金the Natural Science Foundation of Zhejiang Province(No.LQ20F020024)。
文摘Various and intricate varieties of lung disease have made it challenging for computer aided diagnosis to appropriately segment lung lesions utilizing computed tomography(CT)images.This study integrates transfer learning with the attention mechanism to construct a deep learning model that can automatically detect new coronary pneumonia on lung CT images.In this study,using VGG16 pre-trained by ImageNet as the encoder,the decoder was established utilizing the U-Net structure.The attention module is incorporated during each concatenate procedure,permitting the model to concentrate on the critical information and identify the crucial components efficiently.The public COVID-19-CT-Seg-Benchmark dataset was utilized for experiments,and the highest scores for Dice,F1,and Accuracy were 0.9071,0.9076,and 0.9965,respectively.The generalization performance was assessed concurrently,with performance metrics including Dice,F1,and Accuracy over 0.8.The experimental findings indicate the feasibility of the segmentation network proposed in this study.
基金supported by The Shanghai Science and Technology Commission(21S31902700)。
文摘A growing number of skin laser treatments have rapidly evolved and increased their role in the field of dermatology,laser treatment is considered to be used for a variety of pigmentary dermatosis as well as aesthetic problems.The standardized assessment of laser treatment efficacy is crucial for the interpretation and comparison of studies related to laser treatment of skin disorders.In this study,we propose an evaluation method to quantitatively assess laser treatment efficacy based on the image segmentation technology.A tattoo model of Sprague Dawley(SD)rats was established and treated by picosecond laser treatments at varying energy levels.Images of the tattoo models were captured before and after laser treatment,and feature extraction was conducted to quantify the tattooed area and pigment gradation.Subsequently,the clearance rate,which has been a standardized parameter,was calculated.The results indicate that the clearance rates obtained through this quantitative algorithm are comparable and exhibit smaller standard deviations compared with scale scores(4.59%versus 7.93%in the low-energy group,4.01%versus 9.05%in the medium-energy group,and 4.29%versus 10.23%in the high-energy group).This underscores the greater accuracy,objectivity,and reproducibility in assessing treatment responses.The quantitative evaluation of pigment removal holds promise for facilitating faster and more robust assessments in research and development.Additionally,it may enable the optimization of treatments tailored to individual patients,thereby contributing to more effective and personalized dermatological care.
基金supported by Natural Science Foundation Programme of Gansu Province(No.24JRRA231)National Natural Science Foundation of China(No.62061023)Gansu Provincial Science and Technology Plan Key Research and Development Program Project(No.24YFFA024).
文摘Despite its remarkable performance on natural images,the segment anything model(SAM)lacks domain-specific information in medical imaging.and faces the challenge of losing local multi-scale information in the encoding phase.This paper presents a medical image segmentation model based on SAM with a local multi-scale feature encoder(LMSFE-SAM)to address the issues above.Firstly,based on the SAM,a local multi-scale feature encoder is introduced to improve the representation of features within local receptive field,thereby supplying the Vision Transformer(ViT)branch in SAM with enriched local multi-scale contextual information.At the same time,a multiaxial Hadamard product module(MHPM)is incorporated into the local multi-scale feature encoder in a lightweight manner to reduce the quadratic complexity and noise interference.Subsequently,a cross-branch balancing adapter is designed to balance the local and global information between the local multi-scale feature encoder and the ViT encoder in SAM.Finally,to obtain smaller input image size and to mitigate overlapping in patch embeddings,the size of the input image is reduced from 1024×1024 pixels to 256×256 pixels,and a multidimensional information adaptation component is developed,which includes feature adapters,position adapters,and channel-spatial adapters.This component effectively integrates the information from small-sized medical images into SAM,enhancing its suitability for clinical deployment.The proposed model demonstrates an average enhancement ranging from 0.0387 to 0.3191 across six objective evaluation metrics on BUSI,DDTI,and TN3K datasets compared to eight other representative image segmentation models.This significantly enhances the performance of the SAM on medical images,providing clinicians with a powerful tool in clinical diagnosis.
基金support from Strategic Project of Precision Surgery,Tsinghua UniversityInitiative Scientific Research Program,Institute for Intelligent Healthcare,Tsinghua University+5 种基金Tsinghua-Foshan Institute of Advanced ManufacturingNational Natural Science Foundation of China(61735016)Beijing Nova Program(20230484308)Young Elite Scientists Sponsorship Program by CAST(2023QNRC001)Youth Elite Program of Beijing Friendship Hospital(YYQCJH2022-9)Science and Technology Program of Beijing Tongzhou District(KJ2023CX012).
文摘Foundation models(FMs)have rapidly evolved and have achieved signicant accomplishments in computer vision tasks.Specically,the prompt mechanism conveniently allows users to integrate image prior information into the model,making it possible to apply models without any training.Therefore,we proposed a workflow based on foundation models and zero training to solve the tasks of photoacoustic(PA)image processing.We employed the Segment Anything Model(SAM)by setting simple prompts and integrating the model's outputs with prior knowledge of the imaged objects to accomplish various tasks,including:(1)removing the skin signal in three-dimensional PA image rendering;(2)dual speed-of-sound reconstruction,and(3)segmentation ofnger blood vessels.Through these demonstrations,we have concluded that FMs can be directly applied in PA imaging without the requirement for network design and training.This potentially allows for a hands-on,convenient approach to achieving efficient and accurate segmentation of PA images.This paper serves as a comprehensive tutorial,facilitating the mastery of the technique through the provision of code and sample datasets.
基金supported by the National Natural Science(No.U19A2063)the Jilin Provincial Development Program of Science and Technology (No.20230201080GX)the Jilin Province Education Department Scientific Research Project (No.JJKH20230851KJ)。
文摘The visual noise of each light intensity area is different when the image is drawn by Monte Carlo method.However,the existing denoising algorithms have limited denoising performance under complex lighting conditions and are easy to lose detailed information.So we propose a rendered image denoising method with filtering guided by lighting information.First,we design an image segmentation algorithm based on lighting information to segment the image into different illumination areas.Then,we establish the parameter prediction model guided by lighting information for filtering(PGLF)to predict the filtering parameters of different illumination areas.For different illumination areas,we use these filtering parameters to construct area filters,and the filters are guided by the lighting information to perform sub-area filtering.Finally,the filtering results are fused with auxiliary features to output denoised images for improving the overall denoising effect of the image.Under the physically based rendering tool(PBRT)scene and Tungsten dataset,the experimental results show that compared with other guided filtering denoising methods,our method improves the peak signal-to-noise ratio(PSNR)metrics by 4.2164 dB on average and the structural similarity index(SSIM)metrics by 7.8%on average.This shows that our method can better reduce the noise in complex lighting scenesand improvethe imagequality.
基金Natural Science Foundation of Zhejiang Province,Grant/Award Number:LY23F020025Science and Technology Commissioner Program of Huzhou,Grant/Award Number:2023GZ42Sichuan Provincial Science and Technology Support Program,Grant/Award Numbers:2023ZHCG0005,2023ZHCG0008。
文摘Data augmentation plays an important role in training deep neural model by expanding the size and diversity of the dataset.Initially,data augmentation mainly involved some simple transformations of images.Later,in order to increase the diversity and complexity of data,more advanced methods appeared and evolved to sophisticated generative models.However,these methods required a mass of computation of training or searching.In this paper,a novel training-free method that utilises the Pre-Trained Segment Anything Model(SAM)model as a data augmentation tool(PTSAM-DA)is proposed to generate the augmented annotations for images.Without the need for training,it obtains prompt boxes from the original annotations and then feeds the boxes to the pre-trained SAM to generate diverse and improved annotations.In this way,annotations are augmented more ingenious than simple manipulations without incurring huge computation for training a data augmentation model.Multiple comparative experiments on three datasets are conducted,including an in-house dataset,ADE20K and COCO2017.On this in-house dataset,namely Agricultural Plot Segmentation Dataset,maximum improvements of 3.77%and 8.92%are gained in two mainstream metrics,mIoU and mAcc,respectively.Consequently,large vision models like SAM are proven to be promising not only in image segmentation but also in data augmentation.
基金The researchers would like to thank the Deanship of Graduate Studies and Scientific Research at Qassim University for financial support(QU-APC-2025).
文摘Deep learning(DL),derived from the domain of Artificial Neural Networks(ANN),forms one of the most essential components of modern deep learning algorithms.DL segmentation models rely on layer-by-layer convolution-based feature representation,guided by forward and backward propagation.Acritical aspect of this process is the selection of an appropriate activation function(AF)to ensure robustmodel learning.However,existing activation functions often fail to effectively address the vanishing gradient problem or are complicated by the need for manual parameter tuning.Most current research on activation function design focuses on classification tasks using natural image datasets such asMNIST,CIFAR-10,and CIFAR-100.To address this gap,this study proposesMed-ReLU,a novel activation function specifically designed for medical image segmentation.Med-ReLU prevents deep learning models fromsuffering dead neurons or vanishing gradient issues.It is a hybrid activation function that combines the properties of ReLU and Softsign.For positive inputs,Med-ReLU adopts the linear behavior of ReLU to avoid vanishing gradients,while for negative inputs,it exhibits the Softsign’s polynomial convergence,ensuring robust training and avoiding inactive neurons across the training set.The training performance and segmentation accuracy ofMed-ReLU have been thoroughly evaluated,demonstrating stable learning behavior and resistance to overfitting.It consistently outperforms state-of-the-art activation functions inmedical image segmentation tasks.Designed as a parameter-free function,Med-ReLU is simple to implement in complex deep learning architectures,and its effectiveness spans various neural network models and anomaly detection scenarios.
基金Supported by the National Natural Science Foundation of China(No.62001313)the Key Project of Liaoning Provincial Department of Science and Technology(No.2021JH2/10300134,2022JH1/10500004)。
文摘In the realm of medical image segmentation,particularly in cardiac magnetic resonance imaging(MRI),achieving robust performance with limited annotated data is a significant challenge.Performance often degrades when faced with testing scenarios from unknown domains.To address this problem,this paper proposes a novel semi-supervised approach for cardiac magnetic resonance image segmentation,aiming to enhance predictive capabilities and domain generalization(DG).This paper establishes an MT-like model utilizing pseudo-labeling and consistency regularization from semi-supervised learning,and integrates uncertainty estimation to improve the accuracy of pseudo-labels.Additionally,to tackle the challenge of domain generalization,a data manipulation strategy is introduced,extracting spatial and content-related information from images across different domains,enriching the dataset with a multi-domain perspective.This papers method is meticulously evaluated on the publicly available cardiac magnetic resonance imaging dataset M&Ms,validating its effectiveness.Comparative analyses against various methods highlight the out-standing performance of this papers approach,demonstrating its capability to segment cardiac magnetic resonance images in previously unseen domains even with limited annotated data.
基金supported by the National Natural Science Foundation of China(Grant Nos.62072250,61772281,61702235,U1636117,U1804263,62172435,61872203 and 61802212)the Zhongyuan Science and Technology Innovation Leading Talent Project of China(Grant No.214200510019)+3 种基金the Suqian Municipal Science and Technology Plan Project in 2020(S202015)the Plan for Scientific Talent of Henan Province(Grant No.2018JR0018)the Opening Project of Guangdong Provincial Key Laboratory of Information Security Technology(Grant No.2020B1212060078)the Priority Academic Program Development of Jiangsu Higher Education Institutions(PAPD)Fund.
文摘Medical image segmentation,i.e.,labeling structures of interest in medical images,is crucial for disease diagnosis and treatment in radiology.In reversible data hiding in medical images(RDHMI),segmentation consists of only two regions:the focal and nonfocal regions.The focal region mainly contains information for diagnosis,while the nonfocal region serves as the monochrome background.The current traditional segmentation methods utilized in RDHMI are inaccurate for complex medical images,and manual segmentation is time-consuming,poorly reproducible,and operator-dependent.Implementing state-of-the-art deep learning(DL)models will facilitate key benefits,but the lack of domain-specific labels for existing medical datasets makes it impossible.To address this problem,this study provides labels of existing medical datasets based on a hybrid segmentation approach to facilitate the implementation of DL segmentation models in this domain.First,an initial segmentation based on a 33 kernel is performed to analyze×identified contour pixels before classifying pixels into focal and nonfocal regions.Then,several human expert raters evaluate and classify the generated labels into accurate and inaccurate labels.The inaccurate labels undergo manual segmentation by medical practitioners and are scored based on a hierarchical voting scheme before being assigned to the proposed dataset.To ensure reliability and integrity in the proposed dataset,we evaluate the accurate automated labels with manually segmented labels by medical practitioners using five assessment metrics:dice coefficient,Jaccard index,precision,recall,and accuracy.The experimental results show labels in the proposed dataset are consistent with the subjective judgment of human experts,with an average accuracy score of 94%and dice coefficient scores between 90%-99%.The study further proposes a ResNet-UNet with concatenated spatial and channel squeeze and excitation(scSE)architecture for semantic segmentation to validate and illustrate the usefulness of the proposed dataset.The results demonstrate the superior performance of the proposed architecture in accurately separating the focal and nonfocal regions compared to state-of-the-art architectures.Dataset information is released under the following URL:https://www.kaggle.com/lordamoah/datasets(accessed on 31 March 2025).
