Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- David DoermannSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works526 from public data
- 1,645
- 623
- 546
- A global benchmark of algorithms for segmenting the left atrium from late gadolinium-enhanced cardiac magnetic resonance imaging517
This large-scale benchmarking study makes a significant step towards much-improved segmentation methods for atrial LGE-MRIs, and will serve as an important benchmark for evaluating and comparing the future works in the field.
- Large language models for generative information extraction: a survey507
This study surveys the most recent advancements in generative Large Language Models efforts for IE tasks and identifies several insights in technique and promising research directions that deserve further exploration in future studies.
- Combo loss: Handling input and output imbalance in multi-organ segmentation459
A new curriculum learning based loss function that leverages Dice similarity coefficient to deter model parameters from being held at bad local minima and at the same time gradually learn better model parameters by penalizing for false positives/negatives using a cross entropy term is introduced.
- Translating and Segmenting Multimodal Medical Volumes with Cycle- and Shape-Consistency Generative Adversarial Network423
This work proposes a generic cross-modality synthesis approach and shows that these goals can be achieved with an end-to-end 3D convolutional neural network (CNN) composed of mutually-beneficial generators and segmentors for image synthesis and segmentation tasks.
- Morphological diversity of single neurons in molecularly defined cell types417
This study demonstrates the crucial need for quantitative description of complete single-cell anatomy in cell-type classification, as single- cell morphological diversity reveals a plethora of ways in which different cell types and their individual members may contribute to the configuration and function of their respective circuits.
- BoxDiff: Text-to-Image Synthesis with Training-Free Box-Constrained Diffusion354
Three spatial constraints are designed and seamlessly integrated into the denoising step of diffusion models, requiring no additional training and massive annotated layout data, and demonstrate that the proposed constraints can control what and where to present in the images while retaining the ability of Diffusion models to synthesize with high fidelity and diverse concept coverage.
- 316
- 314
- 309
- Deep Learning and Convolutional Neural Networks for Medical Image Computing289
The impact of deep learning on automated disease detection and organ and lesion segmentation and the remaining knowledge gaps that must be overcome to achieve clinician-level performance of automated medical image processing systems are discussed.
- PRGC: Potential Relation and Global Correspondence Based Joint Relational Triple Extraction281
A component to predict potential relations is designed, which constrains the following entity extraction to the predicted relation subset rather than all relations; then a relation-specific sequence tagging component is applied to handle the overlapping problem between subjects and objects; finally, a global correspondence component is designed to align the subject and object into a triple with low-complexity.
- 276
- X2CT-GAN: Reconstructing CT From Biplanar X-Rays With Generative Adversarial Networks276
This work proposes to reconstruct CT from two orthogonal X-rays using the generative adversarial network (GAN) framework, resulting in a high-quality CT volume both visually and quantitatively.
- 260
- 235
- A medical multimodal large language model for future pandemics215
A medical multimodal large language model (Med-MLLM) for radiograph representation learning is reported, which can learn broad medical knowledge from unlabelled data and can be used for clinical tasks that involve both visual and textual data.
- 215
- 203
- 203
- Ambiguity-selective consistency regularization for mean-teacher semi-supervised medical image segmentation202
This paper comprehensively introduces and benchmark a family of plug-and-play strategies for ambiguous target selection from the perspectives of entropy, model uncertainty and label noise self-identification, respectively, and incorporates the estimated ambiguity map into the consistency loss to encourage consensus between the two models' predictions in these informative regions.
- 197
- When MOE Meets LLMs: Parameter Efficient Fine-tuning for Multi-task Medical Applications196
A novel parameter efficient fine-tuning framework for multi-task medical applications, dubbed as MOELoRA, which aims to absorb both the benefits of mixture-of-expert (MOE) for multi-task learning and low-rank adaptation (LoRA) for parameter efficient fine-tuning.
- 176
- Rubik’s Cube+: A self-supervised feature learning framework for 3D medical image analysis173
A pretext task, i.e., Rubik's cube+, is proposed, to pre-train 3D neural networks, forcing networks to learn translation and rotation invariant features from the original 3D medical data, and tolerate the noise of the data at the same time.
- Self-supervised Feature Learning for 3D Medical Images by Playing a Rubik’s Cube173
A novel proxy task, i.e., Rubik's cube recovery, is formulated to pre-train 3D neural networks and it is shown that the self-supervised learning approach can substantially boost the accuracy of 3D deep learning networks on the volumetric medical datasets without using extra data.
- 3D Deep Learning for Efficient and Robust Landmark Detection in Volumetric Data171
An efficient and robust deep learning algorithm capable of full 3D detection in volumetric data is proposed and small 3D patches from a multi-resolution image pyramid are extracted to mitigate the over-fitting issue, thereby increasing detection robustness.
- Deep similarity learning for multimodal medical images170
A novel deep similarity learning method that trains a binary classifier to learn the correspondence of two image patches to show the advantage of the proposed metric for a highly accurate and robust similarity measure.
- 168
- 165
- LLM-ESR: Large Language Models Enhancement for Long-tailed Sequential Recommendation161
A dual-view modeling framework is designed that combines semantics from LLMs and collaborative signals from conventional SRS to address the long-tail item challenge and a retrieval augmented self-distillation method to enhance user preference representation using more informative interactions from similar users.
- Preoperative identification of microvascular invasion in hepatocellular carcinoma by XGBoost and deep learning155
Predictive models using eXtreme Gradient Boosting and deep learning based on CT images and 3D-CNN showed considerable efficacy in identifying MVI preoperatively and may facilitate decision-making in HCC treatment but requires further validation.
- Semi-supervised Credit Card Fraud Detection via Attribute-Driven Graph Representation147
The proposed semi-supervised graph neural network, namely GTAN, outperforms other state-of-the-art baselines on three fraud detection datasets and model the fraud patterns through risk propagation among transactions.
- 142
- Self-supervised learning for medical image data with anatomy-oriented imaging planes136
Thorough experiments on two anatomical structures and representative target tasks demonstrate that the proposed pretext tasks are effective in pretraining deep networks for remarkably boosted performance on the target tasks, and superior to other recent approaches.
- Dynamic Joint Domain Adaptation Network for Motor Imagery Classification133
A novel dynamic joint domain adaptation network based on adversarial learning strategy to learn domain-invariant feature representation, and thus improve EEG classification performance in the target domain by leveraging useful information from the source session is proposed.
- 132
- 132
- 131
- 129
- Comparing to Learn: Surpassing ImageNet Pretraining on Radiographs by Comparing Image Representations128
The experimental results on radiographs show that C2L can outperform ImageNet pretraining and previous state-of-the-art approaches significantly, and is called as Comparing to Learn (C2L) because it learns robust features by comparing different image representations.
- Prototypical Graph Contrastive Learning126
This article proposes a prototypical graph contrastive learning approach, which models the underlying semantic structure of the graph data via clustering semantically similar graphs into the same group and simultaneously encourages the clustering consistency for different augmentations of the same graph.
- Spine detection in CT and MR using iterated marginal space learning125
A novel method is presented that combines Marginal Space Learning (MSL), a recently introduced concept for efficient discriminative object detection, with a generative anatomical network that incorporates relative pose information for the detection of multiple objects to simultaneously detect and label the spinal disks.
- 122
- Self-Loop Uncertainty: A Novel Pseudo-Label for Semi-supervised Medical Image Segmentation121
A novel pseudo-label (namely self-loop uncertainty), generated by recurrently optimizing the neural network with a self-supervised task, is adopted as the ground-truth for the unlabeled images to augment the training set and boost the segmentation accuracy.
- Deep Symmetric Adaptation Network for Cross-Modality Medical Image Segmentation115
This paper presents a novel deep symmetric architecture of UDA for medical image segmentation, which consists of a segmentation sub-network, and two symmetric source and target domain translation sub-networks, and introduces a bidirectional alignment scheme via a shared encoder and two private decoders to simultaneously align features.
- Iterative Multi-domain Regularized Deep Learning for Anatomical Structure Detection and Segmentation from Ultrasound Images109
This paper proposes a multi-domain regularized deep learning method that achieves a superior detection and segmentation accuracy, outperforming other methods by a significant margin and demonstrating competitive capability even compared to human performance.
- All-Around Real Label Supervision: Cyclic Prototype Consistency Learning for Semi-Supervised Medical Image Segmentation108
A novel cyclic prototype consistency learning (CPCL) framework, which is constructed by a labeled-to-unlabeled (L2U) prototypical forward process and an unlabeled- to- labels (U2L) backward process, which synergistically enhance the segmentation network by encouraging morediscriminative and compact features.
- Towards cross-modal organ translation and segmentation: A cycle- and shape-consistent generative adversarial network104
This work systemically analyzes the effect of synthetic data on segmentation and proposes a generic cross‐modality synthesis approach with an end‐to‐end 2D/3D convolutional neural network (CNN) composed of mutually‐beneficial generators and segmentors for image synthesis and segmentation tasks.
- Multi-modal Contrastive Representation Learning for Entity Alignment101
Different from previous works, MCLEA considers task-oriented modality and models the inter-modal relationships for each entity representation, and outperforms state-of-the-art baselines on public datasets under both supervised and unsupervised settings.
- 101
- Development and validation of an artificial intelligence system for grading colposcopic impressions and guiding biopsies99
The CAIADS has potential in assisting beginners and for improving the diagnostic quality of colposcopy and biopsy in the detection of cervical precancer/cancer and demonstrated a superior ability in predicting biopsy sites.
- An evaluation of automatic coronary artery calcium scoring methods with cardiac CT using the orCaScore framework99
An evaluation of five (semi)automatic methods within this framework shows that automatic per patient CVD risk categorization is feasible and CAC lesions at ambiguous locations such as the coronary ostia remain challenging, but their detection had limited impact on CVDrisk determination.
- 99
- 97
- 96
- 95
- 89
- Emerging drug interaction prediction enabled by a flow-based graph neural network with biomedical network87
EmerGNN, a method to predict interactions for emerging drugs, may improve patient care and drug development by providing insight into the effects of using biomedical networks in interaction predictions.
- MedDG: An Entity-Centric Medical Consultation Dataset for Entity-Aware Medical Dialogue Generation86
This work makes the first attempt to build and release a large-scale high-quality Medical Dialogue dataset related to 12 types of common Gastrointestinal diseases named MedDG, with more than 17K conversations collected from the online health consultation community.
- LLMEmb: Large Language Model Can Be a Good Embedding Generator for Sequential Recommendation84
This paper introduces LLMEmb, a novel method leveraging LLM to generate item embeddings that enhance SRS performance and proposes a Supervised Contrastive Fine-Tuning (SCFT) approach to bridge the gap between general-purpose LLM and the recommendation domain.
- InDuDoNet+: A deep unfolding dual domain network for metal artifact reduction in CT images84
Comprehensive experiments on synthesized data and clinical data substantiate the superiority of the proposed methods as well as the superior generalization performance beyond the current state-of-the-art (SOTA) MAR methods.
- Automatic Aorta Segmentation and Valve Landmark Detection in C-Arm CT: Application to Aortic Valve Implantation84
This paper presents a part-based aorta segmentation approach, which can handle aorti structure variation in case that the aortic arch and descending aorte are missing in the volume and is robust under variations of contrast agent.
- Cross-modal coherent registration of whole mouse brains83
A cross-modal registration method is introduced, mBrainAligner, which uses coherent landmark mapping and deep neural networks to align whole mouse brain images to the standard Allen Common Coordinate Framework atlas and generates a whole-brain map of three-dimensional single-neuron morphology and neuron cell types.
- Revisiting Rubik’s Cube: Self-supervised Learning with Volume-Wise Transformation for 3D Medical Image Segmentation83
The experimental results show that the self-supervised learning method can significantly improve the accuracy of 3D deep learning networks on volumetric medical datasets without the use of extra data.
- InDuDoNet: An Interpretable Dual Domain Network for CT Metal Artifact Reduction81
This work builds a joint spatial and Radon domain reconstruction model and utilizes the proximal gradient technique to design an iterative algorithm for solving it, which combines the advantages of model-driven and data-driven methodologies and improves the interpretablility of the framework.
- Detection, Grading and Classification of Coronary Stenoses in Computed Tomography Angiography81
This paper proposes a fast and fully automatic system that is capable of detecting, grading and classifying coronary stenoses in CCTA caused by all types of plaques and shows state-of-the-art performance of the method.
- Molecular Subtyping of Diffuse Gliomas Using Magnetic Resonance Imaging: Comparison and Correlation Between Radiomics and Deep Learning78
Both the radiomics and DCNN models could preoperatively predict the molecular subtypes of diffuse gliomas, and the latter performed better in most circumstances than the former, indicating deep learning features tend to outperform radiomics features in large cohorts.
- 77
- ViLa-MIL: Dual-scale Vision-Language Multiple Instance Learning for Whole Slide Image Classification76
This work proposes a dual-scale visual descriptive text prompt based on the frozen large language model (LLM) to boost the performance of VLM effectively and introduces a context-guided text decoder to enhance the text features by incorporating the multi-granular image contexts.
- LT-Net: Label Transfer by Learning Reversible Voxel-Wise Correspondence for One-Shot Medical Image Segmentation76
A one-shot segmentation method to alleviate the burden of manual annotation for medical images by resorting to the forward-backward consistency, which is widely used in correspondence problems, and additionally learns the backward correspondences from the warped atlases back to the original atlas.
- Multi-task Neural Networks with Spatial Activation for Retinal Vessel Segmentation and Artery/Vein Classification76
A multi-task deep neural network with spatial activation mechanism that is able to segment full retinal vessel, artery and vein simultaneously, without the pre-requirement of vessel segmentation is proposed.
- Robust and Accurate Coronary Artery Centerline Extraction in CTA by Combining Model-Driven and Data-Driven Approaches76
The automatically segmented chambers are exploited to predict the initial position of the major coronary centerlines and define a vessel-specific region-of-interest (ROI) to constrain the following centerline refinement.
- A semi-supervised multi-task learning framework for cancer classification with weak annotation in whole-slide images72
A semi-supervised multi-task learning (MTL) framework for cancer classification that consists of a backbone feature extractor, two task-specific classifiers, and a weight control mechanism that preserves the sequential relationship of these two tasks and guarantees the error back-propagation from the subtyping task to the CRD task under the MTL framework.
- Multi-Anchor Active Domain Adaptation for Semantic Segmentation72
A novel multi-anchor based active learning strategy is introduced to assist domain adaptation regarding the semantic segmentation task by regularizing the latent representation of the target samples compact around multiple anchors through a novel soft alignment loss, more precise segmentation can be achieved.
- Instance-Aware Self-supervised Learning for Nuclei Segmentation72
A novel self-supervised learning framework to deeply exploit the capacity of widely-used convolutional neural networks (CNNs) on the nuclei instance segmentation task, which involves two sub-tasks that enable neural networks to implicitly leverage the prior-knowledge of nuclei size and quantity and mine the instance-aware feature representations from the raw data.
- 71
- Forecasting the long-term trend of COVID-19 epidemic using a dynamic model70
A new model named Dynamic-Susceptible-Exposed-Infective-Quarantined (D-SEIQ) is proposed, by making appropriate modifications of the Susceptibles model and integrating machine learning based parameter optimization under epidemiological rational constraints, that could accurately forecast the long-term trend of the COVID-19 outbreak.
- A New Algorithm for Detecting Text Line in Handwritten Documents70
This work model text line detection as an image segmentation problem by enhancing text line structure using a Gaussian window, and adopting the level set method to evolve text line boundaries.
- 68
- 68
- 68
- Learning Frequency-Adapted Vision Foundation Model for Domain Generalized Semantic Segmentation66
A novel Frequency-Adapted (FADA) learning scheme to separately tackle the content and style information by frequency tokens throughout the learning process, which is able to stabilize the scene content and eliminate its impact to DGSS.
- 66
- RCDNet: An Interpretable Rain Convolutional Dictionary Network for Single Image Deraining66
A novel deep architecture is built, called rain convolutional dictionary network (RCDNet), which embeds the intrinsic priors of rain streaks and has clear interpretability, which facilitates an easy visualization and analysis of what happens inside the network and why it works well in the inference process.
- Uncertainty-aware domain alignment for anatomical structure segmentation66
An uncertainty-aware domain alignment framework to address the domain shift problem in the cross-domain Unsupervised Domain Adaptation (UDA) task is proposed and an Uncertainty Estimation and Segmentation Module (UESM) is designed to obtain the uncertainty map estimation.
- 65
- Machine learning based vesselness measurement for coronary artery segmentation in cardiac CT volumes64
A machine learning based vesselness is proposed by exploiting the rich domain specific knowledge embedded in an expert-annotated dataset and outperforms the conventional Hessian vesselness in both speed and accuracy.
- Guiding the Growth: Difficulty-Controllable Question Generation through Step-by-Step Rewriting63
This work proposes a novel framework that progressively increases question difficulty through step-by-step rewriting under the guidance of an extracted reasoning chain and argues that Question Generation (QG) systems should have stronger control over the logic of generated questions.
- Marginal Space Learning for Efficient Detection of 2D/3D Anatomical Structures in Medical Images63
This paper applies marginal space learning to 2D object detection and performs a thorough comparison between MSL and the alternative full space learning (FSL) approach, showing MSL outperforms FSL in both speed and accuracy.
- 61
- INPREM: An Interpretable and Trustworthy Predictive Model for Healthcare60
The proposed INPREM is designed as a linear model for interpretability while encoding non-linear relationships into the learning weights for modeling the dependencies between and within each visit and outperforms existing approaches with a significant margin.
- Nuclei segmentation with point annotations from pathology images via self-supervised learning and co-training59
A weakly-supervised learning method for nuclei segmentation that only requires point annotations for training is proposed and visual and quantitative results demonstrate the superiority of the method to the state-of-the-art methods, and its competitive performance compared to the fully-super supervised methods.
- System to Guide Transcatheter Aortic Valve Implantations Based on Interventional C-Arm CT Imaging59
A prototypical system based on the ability to acquire a 3D C-arm CT image during transcatheter aortic valve implantations that supports the physician in measuring critical anatomical parameters, finding an optimum C- arm angulation, and guiding the positioning and deployment of the prosthesis by 3D overlay with fluoroscopic images is described.
- Noisy Labels are Treasure: Mean-Teacher-Assisted Confident Learning for Hepatic Vessel Segmentation57
This work proposes a novel mean-teacher-assisted confident learning framework to robustly exploit the noisy labeled data for the challenging hepatic vessel segmentation task, with the adapted confident learning assisted by a third party providing productive guidance.
- Learning Spectral-Decomposited Tokens for Domain Generalized Semantic Segmentation56
A novel Spectral-dEcomposed Token (SET) learning framework to advance the frontier and develops an attention optimization method to bridge the gap between style-affected representation and static tokens during inference.
- A Unified Framework for Generalized Low-Shot Medical Image Segmentation With Scarce Data56
This work proposes a unified framework for generalized low-shot (one- and few-shot) medical image segmentation based on distance metric learning (DML), and proposes adaptive mixing coefficients for the multimodal mixture distributions to adaptively emphasize the modes better suited to the current input.
- Adaptive Convolutional Dictionary Network for CT Metal Artifact Reduction54
This work proposes an adaptive convolutional dictionary network (ACDNet), which leverages both model-based and learning-based methods, and embeds the prior structure of metal artifacts into a deep network, resulting in a clear interpretability for the MAR task.
- GREEN: a Graph REsidual rE-ranking Network for Grading Diabetic Retinopathy54
A Graph REsidual rE-ranking Network (GREEN) is proposed to introduce a class dependency prior into the original image classification network to augments image classification pipeline by re-ranking classification results in a residual aggregation manner.
- Coronary Centerline Extraction via Optimal Flow Paths and CNN Path Pruning54
The robustness and stability of the method are enhanced by using a model-based detection of coronary specific territories and main branches to constrain the search space and a convolutional neural network classifier for removing extraneous paths in the detected centerlines.
- Large Language Model Distilling Medication Recommendation Model53
This paper introduces a novel approach called Large Language Model Distilling Medication Recommendation (LEADER), which transfers the LLM's proficiency to a more compact model and develops a feature-level knowledge distillation technique, which transfers the LLM's proficiency to a more compact model.
- Learning from multiple annotators for medical image segmentation53
This paper extends the preliminary version of the method presented at the NeurIPS Thirty-fourth Annual Conference on Neural Information Processing Systems by extensively evaluating the model on a new created real-world multiple sclerosis lesion dataset.
- OntoEA: Ontology-guided Entity Alignment via Joint Knowledge Graph Embedding52
An ontology-guided entity alignment method named OntoEA, where both KGs and their ontologies are jointly embedded, and the class hierarchy and theclass disjointness are utilized to avoid false mappings is proposed.
- 49
- 48
- Learning Generalized Medical Image Segmentation from Decoupled Feature Queries46
The proposed decoupled fea- ture query (DFQ) scheme can be seamlessly integrate into the Transformer segmentation model in an end-to-end manner and notably outperforming the runner-up by 1.31% and 1.98% with DSC metric on generalized fundus and prostate benchmarks, respectively.
- An Industry Evaluation of Embedding-based Entity Alignment46
This study evaluates state-of-the-artEmbedding-based entity alignment methods in an industrial context, where the impact of seed mappings with different sizes and different biases is explored.
- 46
- 46
- Large language models leverage external knowledge to extend clinical insight beyond language boundaries45
By synergizing medical knowledge through in-context learning, LLMs can extend clinical insight beyond language barriers in healthcare, significantly reducing language-related disparities of LLM applications and ensuring global benefit in this field.
- 45
- Unsupervised Knowledge Graph Alignment by Probabilistic Reasoning and Semantic Embedding45
This study proposes an iterative framework named PRASE which is based on probabilistic reasoning and semantic embedding, and learns the KG embeddings via entity mappings from a probabilism reasoning system named PARIS, and feeds the resultant entities mappings andembeddings back into PARIS for augmentation.
- Attentive CT Lesion Detection Using Deep Pyramid Inference with Multi-scale Booster45
A Multi-Scale Booster with channel and spatial attention integrated into the backbone Feature Pyramid Network (FPN) that performs superiorly against state-of-the-art approaches on lesion detection.
- Detection of 3D Spinal Geometry Using Iterated Marginal Space Learning45
A novel method that combines Marginal Space Learning (MSL), a recently introduced concept for efficient discriminative object detection, with a generative anatomical network that incorporates relative pose information for the detection of multiple objects is presented.
- The Segmentation and Identification of Handwriting in Noisy Document Images45
This paper presents an approach to the problem of segmenting and identifying handwritten annotations in noisy document images, and finds that the reliable results can be achieved at the word level with a classification accuracy of 97.3%.
- Multi-Source Domain Adaptation by Causal-Guided Adaptive Multimodal Diffusion Networks44
This paper proposes a novel MSDA framework, called Causal-guided Adaptive Multimodal Diffusion Networks (C-AMDN), which incorporates a diffusive adversarial generation model for high-fidelity, efficient adaptation among source and target domains, along with deep causal inference re-weighting mechanism for the decision-making process.
- 44
- TW-GAN: Topology and width aware GAN for retinal artery/vein classification44
A novel Topology and Width Aware Generative Adversarial Network is proposed, which, for the first time, integrates the topology connectivity and vessel width information into the deep learning framework for A/V classification.
- Domain Adaptation Meets Zero-Shot Learning: An Annotation-Efficient Approach to Multi-Modality Medical Image Segmentation44
This work proposes a new paradigm of ZSL specifically for medical images utilizing cross-modality information and develops an inheritance attention module to recalibrate the relation prototypes to enhance the inheritance process.
- Brain-wide single neuron reconstruction reveals morphological diversity in molecularly defined striatal, thalamic, cortical and claustral neuron types44
This work reconstructed and analyzed the complete morphologies of 1,708 neurons from the striatum, thalamus, cortex and claustrum and classified these cells into multiple morphological and projection types and identified a set of region-specific organizational rules of long-range axonal projections at the single cell level.
- 44
- Multi-Modality Generative Adversarial Networks with Tumor Consistency Loss for Brain MR Image Synthesis43
A multi-modality generative adversarial network (MGAN) is proposed to synthesize three high-quality MR modalities from one MR modality T2 simultaneously, to preserve the critical tumor information in the generated modalities.
- 43
- 41
- Seg4Reg Networks for Automated Spinal Curvature Estimation41
A new pipeline to perform accurate spinal curvature estimation is proposed, named as Seg4Reg, which contains two deep neural networks focusing on segmentation and regression, respectively, and achieves 21.71 SMAPE in the testing set.
- Deep Reinforcement Learning for Vessel Centerline Tracing in Multi-modality 3D Volumes41
This work proposes a unified end-to-end deep reinforcement learning approach for robust vessel centerline tracing in multi-modality 3D medical volumes to learn an artificial agent to interact with surrounding environment and collect rewards from the interaction.
- Review of Deep Learning Methods in Mammography, Cardiovascular, and Microscopy Image Analysis41
A snapshot of this fast growing field specifically for mammography, cardiovascular, and microscopy image analysis is provided, briefly explain the popular deep neural networks and summarize current deep learning achievements in various tasks such as detection, segmentation, and classification in these heterogeneous imaging modalities.
- Fast Automatic Detection of Calcified Coronary Lesions in 3D Cardiac CT Images41
This work proposes a novel learning-based, fully automatic algorithm for detection of calcified lesions in contrast-enhanced CT data that is quite robust to the estimates of the centerline and works well in practice.
- Deep Image Clustering with Category-Style Representation40
This paper proposes a novel deep image clustering framework to learn a category-style latent representation in which the category information is disentangled from image style and can be directly used as the cluster assignment.
- 39
- OctopusNet: A Deep Learning Segmentation Network for Multi-modal Medical Images39
This work proposes a novel deep learning architecture, namely OctopusNet, to better leverage and fuse the information contained in multi-modalities and demonstrates that the framework outperforms the commonly-used feature fusion approaches and yields the state-of-the-art segmentation accuracy.
- Pyramid Network with Online Hard Example Mining for Accurate Left Atrium Segmentation39
A deep neural network based solution for automated left atrium segmentation in gadolinium-enhanced MR volumes with promising performance and an Online Hard Negative Example Mining strategy to identify voxels in slices with low classification certainties and penalize the wrong predictions on them is proposed.
- 38
- 38
- Seg4Reg+: Consistency Learning Between Spine Segmentation and Cobb Angle Regression38
This study proposes a novel multi-task framework, named Seg4Reg+, which jointly optimizes the segmentation and regression networks and designs a novel triangle consistency learning to train the two networks jointly for global optimization.
- Precise Lumen Segmentation in Coronary Computed Tomography Angiography38
This work proposes a novel, model-guided segmentation approach based on a Markov random field formulation with convex priors which assures the preservation of the tubular structure of the coronary lumen and yields very accurate lumen segmentations even in the presence of calcified and non-calcified plaques.
- NTIRE 2025 Challenge on Day and Night Raindrop Removal for Dual-Focused Images: Methods and Results37
This paper reviews the NTIRE 2025 Challenge on Day and Night Raindrop Removal for Dual-Focused Images to establish a new and powerful benchmark for the task of removing raindrops under varying lighting and focus conditions.
- Interactive Segmentation as Gaussian Process Classification37
The main specificities of the proposed GPCIS lie in: 1) Under the explicit guidance of the derived GP posterior, the information contained in clicks can be finely propagated to the entire image and then boost the segmentation; 2) the accuracy of predictions at clicks has good theoretical support.
- DS-Depth: Dynamic and Static Depth Estimation via a Fusion Cost Volume37
A novel dynamic cost volume that exploits residual optical flow to describe moving objects, improving incorrectly occluded regions in static cost volumes used in previous work and demonstrating that the model outperforms previously published baselines for self-supervised monocular depth estimation.
- 37
- Leveraging Multiple Types of Domain Knowledge for Safe and Effective Drug Recommendation37
The proposed CSEDrug enhances the drug encoding and DDIs controlling by leveraging multi-faceted drug knowledge, including molecule structures of drugs, Synergistic DDIs (SDDIs), and AntagonisticDDIs (ADDIs), by a graph-based drug encoder and multiple loss functions.
- Self-Supervised CycleGAN for Object-Preserving Image-to-Image Domain Adaptation37
The experimental results demonstrate that the proposed OP-GAN can yield visually plausible translated images and significantly improve the semantic segmentation accuracy in domain adaptation scenarios with off-the-shelf deep learning networks such as PSPNet and U-Net.
- 36
- Precise Segmentation of Multiple Organs in CT Volumes Using Learning-Based Approach and Information Theory36
The proposed method can be applied to CT images to provide visual guidance to physicians during the computer-aided diagnosis, treatment planning and image-guided radiotherapy to treat cancers in pelvic region.
- MedJourney: Benchmark and Evaluation of Large Language Models over Patient Clinical Journey35
A typical patient’s clinical journey is divided into four stages: planning, access, delivery and ongoing care, resulting in a comprehensive benchmark comprising 12 datasets, of which five are newly introduced, and seven are constructed from existing datasets.
- 35
- Crossover-Net: Leveraging vertical-horizontal crossover relation for robust medical image segmentation35
A novel deep model termed as Crossover-Net for robust segmentation in medical images, showing promising results compared with the current state-of-the-art methods.
- 35
- 34
- 34
- Using machine learning algorithms for predicting cognitive impairment and identifying modifiable factors among Chinese elderly people34
Doing more leisure activities, doing more gardening work, and engaging in more activities combined were associated with decreased risk of cognitive impairment after 3 years.
- Enquire One’s Parent and Child Before Decision: Fully Exploit Hierarchical Structure for Self-Supervised Taxonomy Expansion34
By better exploiting the hierarchical structure and optimizing taxonomy’s coherence, HEF vastly surpasses the prior state-of-the-art on three benchmark datasets by an average improvement of 46.7% in accuracy and 32.3% in mean reciprocal rank.
- Difficulty-Aware Glaucoma Classification with Multi-rater Consensus Modeling34
A multi-branch model structure is proposed to predict the most sensitive, most specifical and a balanced fused result for the input images in order to encourage the sensitivity branch and specificity branch to generate consistent results for consensus labels and opposite results for disagreement labels.
- 34
- 34
- Rethinking Brain Tumor Segmentation From the Frequency Domain Perspective33
The Harmonized Frequency Fusion Network (HFF-Net) is proposed, which rethinks brain tumor segmentation from a frequency-domain perspective and introduces an Adaptive Laplacian Convolution (ALC) module that adaptively emphasizes critical high-frequency details using dynamically updated convolution kernels.
- TFCD: Towards Multi-modal Sarcasm Detection via Training-Free Counterfactual Debiasing33
A Training-Free Counterfactual Debiasing framework TFCD is proposed, which first formulates the causalities among variables in MSD via a tailored causal graph, and extracts biases from the conventionally-trained model by generating counterfactual utterances and contexts and mitigates them using element-wise subtraction.
- Dual Adversarial Network for Deep Active Learning33
This work investigates the overlapping problem of recent uncertainty-based approaches and proposes a dual adversarial network, namely DAAL, for this purpose, which learns to select the most uncertain and representative data points in one-stage.
- Generative Adversarial Networks for Video-to-Video Domain Adaptation32
Comprehensive experiments show that the domain gap could be substantially narrowed down by the proposed VideoGAN, which has an X-shape generator to preserve the intra-video consistency during translation and a loss function to tune the color distribution of each translated frame.
- TR-GAN: Topology Ranking GAN with Triplet Loss for Retinal Artery/Vein Classification32
A Topology Ranking Generative Adversarial Network (TR-GAN) is proposed to improve the topology connectivity of the segmented arteries and veins, and further to boost the A/V classification performance on the publicly available AV-DRIVE dataset.
- 32
- A Deep Learning Approach to Predict Conductive Hearing Loss in Patients With Otitis Media With Effusion Using Otoscopic Images31
The model in this diagnostic/prognostic study provided greater accuracy in prediction of CHL in ears with OME than those obtained from the logistic regression model and otologists, indicating great potential for the use of artificial intelligence tools to facilitate CHL evaluation when CHL is unable to be measured.
- DeltaNet:Conditional Medical Report Generation for COVID-19 Diagnosis31
DeltaNet is evaluated on a COVID-19 dataset, where DeltaNet outperforms state-of-the-art approaches and its generalization capabilities on the public IU-Xray and MIMIC-CXR datasets for chest-related diseases are validated.
- Pre-trained Online Contrastive Learning for Insurance Fraud Detection30
This paper proposes an innovative online learning method for medical insurance fraud detection, named POCL, which combines contrastive learning pre-training with online updating strategies and has significant advantages in accuracy compared to the state-of-the-art baseline methods, while also exhibiting lower running time and space consumption.
- Relational Experience Replay: Continual Learning by Adaptively Tuning Task-Wise Relationship30
Relational Experience Replay (RER) is proposed, a bi-level learning framework, to adaptively tune task-wise relationships and sample importance within each task to achieve a better ‘stability’ and ‘plasticity’ trade-off.
- 30
- 30
- Sliding Window Attention Training for Efficient Large Language Models29
This paper first attributes the inefficiency of Transformers to the attention sink phenomenon resulting from the high variance of softmax operation, then replaces softmax with the sigmoid function and utilize a balanced ALiBi and Rotary Position Embedding for efficient information compression and retention.
- Cross-Domain Gated Learning for Domain Generalization29
The proposed CDG training strategy can excellently enforce the network to exploit the intrinsic features of objects from the multi-domain data, and achieve a new state-of-the-art domain generalization performance on these benchmarks.
- S-CUDA: Self-cleansing unsupervised domain adaptation for medical image segmentation29
This paper identifies a novel problem hidden in medical image segmentation, which is unsupervised domain adaptation on noisy labeled data, and proposes a novel algorithm named "Self-Cleansing Unsupervised Domain Adaptation" (S-CDUA) to address such issue.
- Single-character type identification29
This paper addresses the problem of character type identification independent of its content, including handwritten/printed Chinese character identification and printed Chinese/English character identification, based on only one character, by exploiting some effective features of OCR technologies.
- Prototype Correlation Matching and Class- Relation Reasoning for Few-Shot Medical Image Segmentation28
A prototype correlation matching module to mine representative prototypes that can characterize diverse visual information of different appearances well and a class-relation reasoning module to segment unseen novel medical objects via reasoning inter-class relations between base and novel classes are proposed.
- Face Completion with Semantic Knowledge and Collaborative Adversarial Learning28
This work proposes a collaborative adversarial learning approach to face completion to explicitly induce the training process, formulated under a novel generative framework called collaborative GAN (collaGAN), which allows better semantic understanding of a target object through collaborative learning of multiple tasks including face completion, landmark detection, and semantic segmentation.
- 28
- Editing Factual Knowledge and Explanatory Ability of Medical Large Language Models27
This work proposes MedLaSA, a novel Layer-wise Scalable Adapter strategy for medical model editing that harnesses the strengths of both adding extra parameters and locate-then-edit methods for medical model editing and incorporates scalable adapters into the dense layers of LLMs.
- QUBIQ: Uncertainty Quantification for Biomedical Image Segmentation Challenge27
The obtained results indicate the importance of the ensemble models, as well as the need for further research to develop efficient 3D methods for uncertainty quantification methods in 3D segmentation tasks.
- CTNeRF: Cross-time Transformer for dynamic neural radiance field from monocular video27
This work proposes a novel approach that builds upon a recent generalization NeRF, which aggregates nearby views onto new viewpoints to aggregate the features of object motion and introduces a module that operates in both the time and frequency domains to aggregate the features of object motion.
- MRL-Seg: Overcoming Imbalance in Medical Image Segmentation With Multi-Step Reinforcement Learning27
This work proposes a novel approach based on multi-step reinforcement learning, which integrates prior knowledge of medical images and pixel-wise segmentation difficulty into the reward function, and shows that it surpasses other state-of-the-art methods in highly imbalanced scenarios.
- 27
- Lower-extremity fatigue fracture detection and grading based on deep learning models of radiographs27
The deep learning–based models could be applied to the radiological diagnosis of plain radiographs for assisting in the detection and grading of fatigue fractures on tibiofibula and foot.
- Pairwise learning for medical image segmentation27
A conjugate fully convolutional network (CFCN) where pairwise samples are input for capturing a rich context representation and guide each other with a fusion module to avoid the overfitting problem introduced by intra-class heterogeneity and boundary ambiguity with a small number of training samples is proposed.
- Leveraging Undiagnosed Data for Glaucoma Classification with Teacher-Student Learning27
Experiments show that the proposed framework is able to utilize the undiagnosed data effectively to improve the glaucoma prediction performance.
- 26
- Combinatorial CNN-Transformer Learning with Manifold Constraints for Semi-supervised Medical Image Segmentation26
A novel CNN-Transformer learning framework in the manifold space for semi-supervised medical image segmentation that outperforms previous state-of-the-art methods on three public medical image segmentation benchmarks.
- UniHead: Unifying Multi-Perception for Detection Heads26
The UniHead proposes a dual-axial aggregation transformer (DAT) to adeptly model long-range dependencies, thereby achieving GP, and devises a cross-task interaction transformer (CIT) that facilitates interaction between the classification and localization branches, thus aligning the two tasks.
- 26
- 26
- GRAND: A large-scale dataset and benchmark for cervical intraepithelial Neoplasia grading with fine-grained lesion description26
A colposcopic image dataset for GRAding cervical intraepithelial neoplasia with fine-grained lesion Description (GRAND), and a human-machine confrontation to evaluate the potential of the proposed benchmark framework for clinical applications.
- MI$$^2$$GAN: Generative Adversarial Network for Medical Image Domain Adaptation Using Mutual Information Constraint26
A novel GAN is proposed (namely MI$^2$GAN) to maintain image-contents during cross-domain I2I translation and significantly improve the generalization performance of widely used deep learning networks (e.g., U-Net).
- Deep Learning Based Automatic Segmentation of Pathological Kidney in CT: Local Versus Global Image Context26
This work proposes to combine deep learning and marginal space learning (MSL), both using local context, for robust kidney detection and segmentation, and eliminates all gross failures (i.e., segmentation is totally off) in a direct application of MSL.
- Computer Aided Diagnosis Using Multilevel Image Features on Large-Scale Evaluation26
This paper presents a principled three-tiered image feature learning approach to capture task specific and data-driven class discriminative statistics from an annotated image database that integrates voxel-, instance-, and database-level feature learning, aggregation and parsing.
- 26
- Fast and Robust 3-D MRI Brain Structure Segmentation26
A novel method for the automatic detection and segmentation of (sub-)cortical gray matter structures in 3-D magnetic resonance images of the human brain using the recently introduced concept of Marginal Space Learning (MSL).
- AnomalyXFusion: Multi-modal Anomaly Synthesis with Diffusion25
The AnomalyXFusion framework, designed to harness multi-modality information to enhance the quality of synthesized abnormal samples, is presented and a new dataset is proposed, called MVTec Caption, to reveal the multi-modality representational power of AnomalyXFusion.
- Probing the Impacts of Visual Context in Multimodal Entity Alignment25
This work proposes to learn entity representations from graph structures and visual context, and combine feature similarities to find alignments at the output level, and explores a mechanism which utilizes classification techniques and entity types to remove potentially un-helpful images during alignment learning and inference.
- 25
- 25
- Cross-modality Neuroimage Synthesis: A Survey24
A comprehensive review of cross-modality synthesis for neuroimages, from the perspectives of weakly supervised and unsupervised settings, loss functions, evaluation metrics, imaging modalities, datasets, and downstream applications based on synthesis.
- 24
- Exploring Social Media for Early Detection of Depression in COVID-19 Patients24
This paper investigated the relationship between COVID-19 infection and depression through social media analysis and proposed a deep neural network for early prediction of depression risk, which has the potential to enable public health organizations to initiate prompt intervention with high-risk patients.
- Superpixel-Guided Label Softening for Medical Image Segmentation24
Experimental results on a brain MRI dataset and an optical coherence tomography dataset demonstrate that this conceptually simple and implementation-wise easy method achieves overall superior segmentation performances to baseline and comparison methods for both 3D and 2D medical images.
- Improving accuracy in coronary lumen segmentation via explicit calcium exclusion, learning-based ray detection and surface optimization24
A novel fully-automatic method to accurately segment the lumen of coronary arteries in the presence of calcified and non-calcified plaque is presented, which is comparable to clinical experts and superior to current automatic methods.
- Fast and Automatic Heart Isolation in 3D CT Volumes: Optimal Shape Initialization24
Marginal space learning (MSL) is used to efficiently estimate the position, orientation, and scale of the heart and post-processing is exploited to exclude the rib cage from the heart mask.
- InMu-Net: Advancing Multi-modal Intent Detection via Information Bottleneck and Multi-sensory Processing23
This paper proposes InMu-Net, a simple yet effective framework for MID from the Information bottleneck and Multi-sensory processing perspective, and introduces a saliency preservation loss to prevent the dropping of intent-relevant information.
- Double-Uncertainty Guided Spatial and Temporal Consistency Regularization Weighting for Learning-Based Abdominal Registration23
A mean-teacher based registration framework, which incorporates an additional temporal consistency regularization term by encouraging the teacher model's prediction to be consistent with that of the student model, which can promisingly advance the original learning-based method in terms of efficient hyperparameter tuning and a better tradeoff between accuracy and smoothness.
- 23
- A discriminative model-constrained EM approach to 3D MRI brain tissue classification and intensity non-uniformity correction23
A fully automated method for tissue classification, which is the segmentation into cerebral gray matter, cerebral white matter, and cerebral spinal fluid, and intensity non-uniformity (INU) correction in brain magnetic resonance imaging (MRI) volumes, which combines supervised MRI modality-specific discriminative modeling and unsupervised statistical expectation maximization (EM) segmentations into an integrated Bayesian framework.
- Multi-part Left Atrium Modeling and Segmentation in C-Arm CT Volumes for Atrial Fibrillation Ablation23
A part based LA model is proposed (including the chamber, appendage, and four major PVs) and each part is a much simpler anatomical structure compared to the holistic one and a statistical shape constraint is proposed during the estimation of pose parameters of different parts.
- 23
- Fighting against Organized Fraudsters Using Risk Diffusion-based Parallel Graph Neural Network22
RDPGL, a novel Risk Diffusion-based Parallel Graph Learning approach, to fighting against medical insurance criminal gangs is proposed, which could detect more organized fraud claims with relatively high precision compared with state-of-the-art baselines.
- MADAv2: Advanced Multi-Anchor Based Active Domain Adaptation Segmentation22
This work proposes to introduce active sample selection to assist domain adaptation regarding the semantic segmentation task and introduces a powerful semi-supervised domain adaptation strategy to alleviate the long-tail distribution problem and further improve the segmentation performance.
- A deep weakly semi-supervised framework for endoscopic lesion segmentation22
A weakly semi-supervised segmentation framework, called Point Segmentation Transformer (Point SEGTR), which largely reduces the demand of pixel-level annotations significantly and proposes two regularization terms, i.e., multi-point consistency and symmetric consistency, to boost the quality of pseudo labels.
- QEN: Applicable Taxonomy Completion via Evaluating Full Taxonomic Relations22
The Quadruple Evaluation Network (QEN) is proposed, a novel taxonomy completion framework that utilizes easily accessible term descriptions as input, and applies pretrained language model and code attention for accurate inference while reducing online computation.
- A Hierarchical Feature Constraint to Camouflage Medical Adversarial Attacks22
A novel hierarchical feature constraint (HFC) is proposed as an add-on to existing adversarial attacks, which encourages the hiding of the adversarial representation within the normal feature distribution and demonstrates the superiority of this adversarial attack method as it bypasses an array of state-of-the-art adversarial detectors more easily.
- Marginal Space Deep Learning: Efficient Architecture for Detection in Volumetric Image Data22
This work proposes Marginal Space Deep Learning MSDL as an effective solution, that combines the strengths of efficient object parametrization in hierarchical marginal spaces with the automated feature design of Deep Learning DL network architectures, and proposes to simplify the network by replacing the standard, pre-determined feature sampling pattern with a sparse, adaptive, self-learned pattern.
- 22
- Federated Modality-Specific Encoders and Multimodal Anchors for Personalized Brain Tumor Segmentation21
This work proposes a new FL framework with federated modality-specific encoders and multimodal anchors (FedMEMA) to simultaneously address the two concurrent issues of inter-modal heterogeneity and outperforms various up-to-date methods for multimodal and personalized FL.
- 21
- Simultaneous alignment and surface regression using hybrid 2D–3D networks for 3D coherent layer segmentation of retinal OCT images with full and sparse annotations21
This work presents a novel framework based on hybrid 2D-3D convolutional neural networks (CNNs) to obtain continuous 3D retinal layer surfaces from OCT volumes, which works well with both full and sparse annotations.
- 21
- A New Bidirectional Unsupervised Domain Adaptation Segmentation Framework21
A bidirectional UDA (BiUDA) framework based on disentangled representation learning for equally competent two-way UDA performances is proposed, which employs a unified domain-aware pattern encoder which not only can adaptively encode images in different domains through a domain controller, but also improve model efficiency by eliminating redundant parameters.
- Noninvasive hemodynamic assessment, treatment outcome prediction and follow‐up of aortic coarctation from MR imaging21
An end-to-end system comprised of shape and computational models, their personalization setup using MR imaging, and a fast, noninvasive method based on computational fluid dynamics (CFD) to estimate the pre- and postoperative hemodynamics for coarctation patients is proposed.
- Learning-Based Detection and Tracking in Medical Imaging: A Probabilistic Approach21
This chapter presents a probabilistic framework that relies on anatomically indexed component-based object models which integrate several sources of information to determine the temporal trajectory of the deformable target and demonstrates various medical image analysis applications with focus on cardiology.
- A collaborative large language model for drug analysis20
A knowledge-grounded collaborative large language model, DrugGPT is proposed to make accurate, evidence-based and faithful recommendations that can be used for clinical decisions and outperforms existing LLMs to answer questions related to drug interactions and recommendations.
- AI-Driven Diagnostic Assistance in Medical Inquiry: Reinforcement Learning Algorithm Development and Validation20
This retrospective analysis of predominantly respiratory pediatric presentations in emergency and pediatrics departments demonstrated that an AI-driven diagnostic assistant had high diagnostic performance both in stand-alone use and in simulated collaboration with clinicians.
- Wearable-based behaviour interpolation for semi-supervised human activity recognition20
This work introduces a deep semi-supervised HAR approach, MixHAR, which concurrently uses labelled and unlabelled activities, and proposes a mixing calibration mechanism to mitigate the activityintrusion problem during mixing.
- Combating Mode Collapse via Offline Manifold Entropy Estimation20
A novel training pipeline to address the mode collapse issue of GANs is proposed to generalize the discriminator as feature embedding and maximize the entropy of distributions in the embedding space learned by the discriminators.
- Distractor-Aware Neuron Intrinsic Learning for Generic 2D Medical Image Classifications20
A novel distractor-aware loss is formulated that encourages large distance between the original image and its distractor in the feature space and is combined with the original classification loss to update network parameters by back-propagation.
- Conquering Data Variations in Resolution: A Slice-Aware Multi-Branch Decoder Network20
A slice-aware 2.5D network that emphasizes extracting discriminative features utilizing not only in-plane semantics but also out-of-plane coherence for each separate slice is proposed to tackle the mismatch between the intra- and inter-slice information.
- 20
- 19
- 19
- mBrainAligner-Web: a web server for cross-modal coherent registration of whole mouse brains19
mBrainAligner-Web, a web server with a user-friendly interface that allows to configure and run mBrainAlignmenter locally or remotely across platforms, is presented.
- 19
- CTA Coronary Labeling through Efficient Geodesics between Trees Using Anatomy Priors19
An efficient realization of recent work on unique geodesic paths between tree shapes for the application of matching coronary arteries to a standard model of coronary anatomy in order to label the coronary arteries is presented.
- Robust Point Matching for Non-Rigid Shapes: A Relaxation Labeling Based Approach19
Experiments demonstrate the effectiveness of the approach: it outperforms shape context and TPS-RPM algorithms under non-rigid deformation and noise on a public data set.
- Safeguarding Fraud Detection from Attacks: A Robust Graph Learning Approach18
This paper designs an attack algorithm tampering with both features and structures of graph data to simulate fraudsters' attacking behaviors in real-life complex fraud scenarios and applies singular value decomposition to the graph and learns the decomposed matrices in a GNN model with specifically designed joint losses.
- 18
- 18
- Lesion Guided Explainable Few Weak-Shot Medical Report Generation18
A lesion guided explainable few weak-shot medical report generation framework that learns correlation between seen and novel classes through visual and semantic feature alignment, aiming to generate medical reports for diseases not observed in training.
- Signature-Based Document Image Retrieval18
This work developed a fully automatic signature-based document image retrieval system that handles automatic detection and segmentation of signatures from document images and Translation, scale, and rotation invariant signature matching for document image retrieved.
- Example based non-rigid shape detection18
This paper proposes a novel machine learning based approach to achieve a refined shape detection result, and trains a model that has the largest response on a reference shape and a smaller response on other shapes.
- Classification-Based Spatial Error Concealment for Visual Communications18
This paper proposes a set of classification-based block concealment schemes, including receiver-side classification, sender- side attachment, and sender-side embedding, and demonstrates that the proposed classification- based error concealment approaches outperform the conventional approaches.
- SP-SLAM: Neural Real-Time Dense SLAM With Scene Priors17
SP-SLAM, a novel neural RGB-D SLAM system that performs tracking and mapping in real-time and introduces an effective optimization strategy for mapping, allowing the system to continuously optimize the poses of all historical input frames during runtime without increasing computational overhead.
- 17
- MCSCSet: A Specialist-annotated Dataset for Medical-domain Chinese Spelling Correction17
This work defines the task of Medical-domain Chinese Spelling Correction (MCSC) and proposes MCSCSet, a large-scale specialist-annotated dataset that contains about 200k samples that offers a medical-domain confusion set consisting of the common error-prone characters in medicine and their corresponding misspellings.
- Quantifying the Effect of Public Activity Intervention Policies on COVID-19 Pandemic Containment Using Epidemiologic Data From 145 Countries17
This study provides more reliable evidence of the quantitative effects of policy interventions on the COVID-19 epidemic and suggested that stricter public activity interventions should be implemented at the early stage of the epidemic for improved containment.
- 17
- 17
- 16
- FemtoDet: An Object Detection Baseline for Energy Versus Performance Tradeoffs16
A new instance boundary enhancement (IBE) module for convolution optimization to overcome the contradiction between the limited capacity of CNNs and detection tasks in diverse spatial representations is developed and a recursive warm-restart (RecWR) for optimizing training strategy to escape the sub-optimization of light-weight detectors by considering the data shift produced in popular augmentations.
- Orientation-Shared Convolution Representation for CT Metal Artifact Learning16
This paper carefully analyze the characteristics of metal artifacts and proposes an orientation-shared convolution representation strategy to adapt the physical prior structures of artifacts, i.e., rotationally symmetrical streaking patterns, which rationally adopts Fourier-series-expansion-based filter parametrization in artifact modeling.
- Automatic quantification of retinal photoreceptor integrity to predict persistent disease activity in neovascular age-related macular degeneration using deep learning16
The integrity of EZ and ELM constituted the strongest predictive factor for PDA in nAMD eyes in real-world clinical practice, and the six-metric model achieved the optimal performance to predict 1-year disease activity.
- Automatic and efficient contrast-based 2-D/3-D fusion for trans-catheter aortic valve implantation (TAVI)16
A fully automatic and efficient system for contrast-based 2-D/3-D fusion for TAVI by integrating the information of aorta segmentation and aortic landmark detection into intensity-based registration, which combines the merits of intensity- based registration and feature/landmark-basedRegistration.
- 16
- LENAS: Learning-Based Neural Architecture Search and Ensemble for 3-D Radiotherapy Dose Prediction15
A novel learning-based ensemble approach named LENAS, which integrates neural architecture search with knowledge distillation for 3-D radiotherapy dose prediction and designs a hybrid loss to optimize the student network, enabling it to recover the knowledge embedded within the teacher networks.
- Adversarial Medical Image With Hierarchical Feature Hiding15
A simple-yet-effective hierarchical feature constraint (HFC), a novel add-on to conventional white-box attacks, which assists to hide the adversarial feature in the target feature distribution and reveals the deficiencies of medical reactive defense and allows to develop more robust defenses in future.
- Identify Event Causality with Knowledge and Analogy15
This paper utilizes a commonsense knowledge graph called ConceptNet to enrich the description of an event sample and reveal the commonalities or associations between different events and retrieves similar events as analogy exams and glean useful experiences from such analogous neigh- bors to better identify the relationship between a new event pair.
- 15
- 15
- 15
- 15
- PASS: Test-Time Prompting to Adapt Styles and Semantic Shapes in Medical Image Segmentation14
A TTA framework called PASS (Prompting to Adapt Styles and Semantic shapes), which jointly learns two types of prompts: the input-space prompt to reformulate the style of the test image to fit into the pretrained model and the semantic-aware prompts to bridge high-level shape discrepancy across domains.
- DFTR: Depth-supervised Fusion Transformer for Salient Object Detection14
Experimental results show that this novel pure Transformer-based SOD framework, namely Depth-supervised hierarchical feature Fusion TRansformer (DFTR), consistently outperforms the existing state-of-the-art methods for both RGB and RGB-D SOD.
- 14
- Evaluation of interpolation methods for surface‐based motion compensated tomographic reconstruction for cardiac angiographic C‐arm data14
The framework of motion estimation using a surface model and motion interpolation to a dense MVF provides the ability for tomographic reconstruction using a motion compensation technique.
- Artificial intelligence-assisted detection of nasopharyngeal carcinoma on endoscopic images: a national, multicentre, model development and validation study13
A deep learning algorithm to enhance an otolaryngologist's diagnostic capabilities by differentiating between nasopharyngeal carcinoma, benign hyperplasia, and normal nasopharynx during endoscopic examination is developed.
- 13
- 13
- 13
- 13
- 13
- 13
- A Layered Debating Multi-Agent System for Similar Disease Diagnosis12
A novel framework that leverages LLMs and a multi-agent system to achieve accurate disease diagnosis through a process of repeated debate and reassessment is probed, demonstrating its potential to enhance diagnostic precision in healthcare.
- 12
- VisorGPT: Learning Visual Prior via Generative Pre-Training12
Experimental results demonstrate that VisorGPT can effectively model the visual prior, which can be employed for many vision tasks, such as customizing accurate human pose for conditional image synthesis models like ControlNet.
- 12
- 12
- Recist-Net: Lesion Detection Via Grouping Keypoints On Recist-Based Annotation12
RecIST-Net is proposed, a new approach to lesion detection in which the four extreme points and center point of the RECIST diameters are detected and achieves a sensitivity of 92.49% at four false positives per image, outperforming other recent methods including those using multi-task learning.
- CONNER: A Cascade Count and Measurement Extraction Tool for Scientific Discourse12
This paper proposes CONNER, a cascade count and measurement extraction tool that can identify entities and the corresponding relations in a two-step pipeline model for SemEval 2021 Task 8: MeasEval.
- Cross-denoising Network against Corrupted Labels in Medical Image Segmentation with Domain Shift12
This paper proposes a novel robust cross-denoising framework using two peer networks to address domain shift and corrupted label problems with a peer-review strategy, and introduces a class-imbalanced cross learning using most confident predictions at class-level.
- A Macro-Micro Weakly-Supervised Framework for AS-OCT Tissue Segmentation12
A novel framework to segment the target tissues accurately for the AS-OCT images, by using the combination of weakly-annotated images (majority) and fully-annotation images (minority), which outperforms the state-of-the-art semi-/weakly- supervised methods and has a comparable performance as the fully-supervised method.
- 12
- Efficient Detection of Native and Bypass Coronary Ostia in Cardiac CT Volumes: Anatomical vs. Pathological Structures12
A new searching strategy is proposed to efficiently guide the focus of analysis and, at the same time, reduce the false positive detections in coronary exam workflow.
- Automatic left ventricle detection in MRI images using marginal space learning and component-based voting12
This paper proposes to use marginal space learning (MSL) to exploit the recent advances in learning discriminative classifiers, and shows that MSL significantly outperforms FSL in both speed and accuracy.
- SAMVSR: Leveraging Semantic Priors to Zone-Focused Mamba for Video Snow Removal11
This work carefully construct the first framework to incorporate SAM-derived semantic priors into video snow removal, called SAMVSR, and proposes a Zone-Focused Mamba module, a novel Mamba-based architecture that restricts its scanning scope to semantically coherent zones, effectively mitigating irrelevant interactions and enhancing temporal-spatial consistency.
- Federated modality-specific encoders and partially personalized fusion decoder for multimodal brain tumor segmentation11
A new FL framework with federated modality-specific encoders and partially personalized multimodal fusion decoders (FedMEPD) to address the two concurrent issues of intermodal heterogeneity and outperforms various up-to-date methods for multimodal and personalized FL.
- Aspects are Anchors: Towards Multimodal Aspect-based Sentiment Analysis via Aspect-driven Alignment and Refinement11
Aspect-driven Alignment and Refinement (ADAR) is introduced, which is a two-stage coarse-to-fine alignment framework that demonstrates the superiority of the model over state-of-the-art performance in the MABSA task.
- The impact of medical insurance and old-age security on the utilization of medical services by the older population with disabilities11
It is recommended to focus on strengthening the support and health management of these people who are unattended to improve the effective use of health services and better meet their needs.
- 11
- FedMed-ATL: Misaligned Unpaired Cross-Modality Neuroimage Synthesis via Affine Transform Loss11
A novel federated self-supervised learning (FedMed) for brain image synthesis is proposed with better stability than other GAN-based algorithms and reduces the demand for deformable registration while encouraging to leverage the misaligned and unpaired data.
- Combating Mode Collapse in GANs via Manifold Entropy Estimation11
A novel training pipeline to address the mode collapse issue of GANs is proposed, which outperforms the GAN baseline, MaF-GAN on CelebA, and surpasses the recent state-of-the-art energy-based model on the ANIME-FACE dataset.
- 11
- Generalized Organ Segmentation by Imitating One-Shot Reasoning Using Anatomical Correlation11
Extensive experiments show that OrganNet can effectively resist the wide variations in organ morphology and produce state-of-the-art results in one-shot segmentation task and even when compared with fully-supervised segmentation models, OrganNet is still able to produce satisfying segmentation results.
- Imperfect also Deserves Reward: Multi-Level and Sequential Reward Modeling for Better Dialog Management11
A multi-level reward modeling approach that factorizes a reward into a three-level hierarchy: domain, act, and slot is proposed that can provide more accurate and explainable reward signals for state-action pairs in dialogs.
- 11
- 11
- Pairwise Semantic Segmentation via Conjugate Fully Convolutional Network11
This paper proposes a conjugate fully convolutional network (CFCN) where pairwise samples are input and synergistically segmented in the network for capturing a rich context representation and shows a significant performance improvement on pathological liver segmentation.
- RSF-Conv: Rotation-and-Scale Equivariant Fourier Parameterized Convolution for Retinal Vessel Segmentation10
A rotation-and-scale equivariant Fourier parameterized convolution (RSF-Conv) specifically for retinal vessel segmentation is proposed and equivariant Fourier parameterized convolution (RSF-Conv) is provided to provide the corresponding equivariance analysis.
- 10
- Samba: Severity-aware Recurrent Modeling for Cross-domain Medical Image Grading10
A Severity-aware Recurrent Modeling (Samba) method is proposed, which outperforms the VMamba baseline by an average accuracy and reconstructs the intermediate features based on learnable severity bases.
- Separated collaborative learning for semi-supervised prostate segmentation with multi-site heterogeneous unlabeled MRI data10
A separated collaborative learning (SCL) framework for semi-supervised prostate segmentation with multi-site unlabeled MRI data is proposed and can effectively generalize SSL on multi-site unlabeled data and significantly outperform other semi-supervised segmentation methods.
- 10
- Semi-Supervised Convolutional Vision Transformer with Bi-Level Uncertainty Estimation for Medical Image Segmentation10
A novel CNN-Transformer learning framework that allows for effective segmentation of medical images by producing complementary and reliable features and pseudo-label with bi-level uncertainty is developed and significantly surpasses existing semi-supervised approaches on two public medical image segmentation datasets.
- Distributionally Robust Sequential Recommnedation10
This paper proposes a robust sequential recommender framework to overcome the potential distribution shift between the training and testing sets, and analyzes the framework by presenting the generalization error bound of the above minimax objective.
- 10
- 10
- Ensembled ResUnet for Anatomical Brain Barriers Segmentation10
A residual block based U-shape network is constructed with a deep encoder and shallow decoder, and the Tversky loss is introduced to address the issue of the class imbalance between different foreground and the background classes.
- Simultaneous Alignment and Surface Regression Using Hybrid 2D-3D Networks for 3D Coherent Layer Segmentation of Retina OCT Images10
A novel framework based on hybrid 2D-3D convolutional neural networks (CNNs) is proposed to obtain continuous 3D retinal layer surfaces from OCT and achieves superior results to state-of-the-art 2D methods in terms of both layer segmentation accuracy and cross-B-scan 3D continuity.
- Structure-Aware Rank-1 Tensor Approximation for Curvilinear Structure Tracking Using Learned Hierarchical Features10
A new tracking method is proposed that encodes the structure prior of CS in the rank-1 tensor approximation tracking framework, and it also uses the learned hierarchical features via a convolutional neural network (CNN) to improve the tracking robustness.
- Marginal Space Learning10
This chapter presents an efficient 3D learning-based object detection method, called Marginal Space Learning, and introduces the steerable features, as a mechanism to search the orientation space, thus avoiding expensive volume/image rotations.
- Graph cuts based left atrium segmentation refinement and right middle pulmonary vein extraction in C-arm CT10
A graph-based method is proposed by exploiting the graph cuts method to refine results from the model-based segmentation and extract right middle pulmonary veins and demonstrate that the proposed graph- based method is effective and efficient to improve the LA segmentation accuracy and Extract right middle PVs.
- Catheter Tracking via Online Learning for Dynamic Motion Compensation in Transcatheter Aortic Valve Implantation10
This paper presents a novel catheter tracking for motion compensation in the Transcatheter Aortic Valve Implantation (TAVI), and presents an adaptive linear discriminant learning method to build a measurement model online to distinguish catheters from background.
- Model-Based Fusion of Multi-modal Volumetric Images: Application to Transcatheter Valve Procedures10
A novel method to fuse pre-operative and intra-operative information by jointly estimating anatomical models from multiple image modalities is proposed and high-quality patient-specific models are integrated into the imaging environment of operating rooms to guide cardiac interventions.
- Left ventricle endocardium segmentation for cardiac CT volumes using an optimal smooth surface10
This paper focuses on improving the LV endocardium segmentation accuracy by searching for an optimal smooth mesh that tightly encloses the whole blood pool by solving the convex quadratic programming problem.
- 9
- How much Medical Knowledge do LLMs have? An Evaluation of Medical Knowledge Coverage for LLMs9
This work proposes a systematic evaluation framework, MedKGEval, to assess the coverage of medical knowledge in LLMs through the lens of medical knowledge graphs, and introduces a novel knowledge-oriented evaluation approach that encompasses the assessment of knowledge coverage across entities, relations, and triples.
- Learning Long-form Video Prior via Generative Pre-Training9
Experimental results validate that the approach has great potential for learning long-form video prior and that it is efficient to employ visual locations like bounding boxes and keypoints to represent key information in videos.
- 9
- Multi-Constraint Transferable Generative Adversarial Networks for Cross-Modal Brain Image Synthesis9
A novel Brain Generative Adversarial Network (BrainGAN) that explores GANs with multi-constraint and transferable property for cross-modal brain image synthesis, and shows how BrainGAN can learn meaningful tissue representations with rich variability of brain images.
- Distributional Fairness-aware Recommendation9
This article defines a novel type of fairness, where it is required that the performance distributions across different user groups should be similar, and proves that with the same performance distribution, the numerical characteristics of the group performance are also the same.
- Automatic view plane prescription for cardiac magnetic resonance imaging via supervision by spatial relationship between views9
A clinic-compatible, annotation-free system for automatic CMR view planning that achieves superior accuracy to existing approaches including conventional atlas-based and newer deep-learning-based in prescribing the four standard CMR planes but also demonstrates prescription of the first cardiac-anatomy-oriented plane from the body-oriented scout.
- 9
- Refining BERT Embeddings for Document Hashing via Mutual Information Maximization9
Experimental results on three benchmark datasets demonstrate that the proposed method is able to generate hash codes that outperform existing ones learned from BOW features by a substantial margin.
- 9
- 9
- Rethinking Facial Expression Recognition in the Era of Multimodal Large Language Models: Benchmark, Datasets, and Beyond8
FERBench is provided, a systematic benchmark that incorporates 20 state-of-the-art MLLMs across four widely used FER datasets and a unified and interpretable FER foundation model termed UniFER-7B is developed, which outperforms many open-sourced and closed-source generalist MLLMs.
- Clinical pathway-aware large language models for reliable and transparent medical dialogue8
By enhancing the reliability, coherence, and transparency of AI-generated responses, this pathway-aware approach bridges the gap between LLMs and real-world clinical workflows, improving the accessibility of high-quality telemedicine services, particularly benefiting underserved populations.
- TRACE: Temporally Reliable Anatomically-Conditioned 3D CT Generation with Enhanced Efficiency8
TRACE is introduced, a framework that generates 3D medical images with spatiotemporal alignment using a 2D multimodal-conditioned diffusion approach, and demonstrates that TRACE effectively balances computational efficiency with preserving anatomical fidelity and spatiotemporal consistency.
- MedKP: Medical Dialogue with Knowledge Enhancement and Clinical Pathway Encoding8
The Medical dialogue with Knowledge enhancement and clinical Pathway encoding (MedKP) framework is introduced, which integrates an external knowledge enhancement module through a medical knowledge graph and an internal clinical pathway encoding via medical entities and physician actions.
- Hybrid unsupervised representation learning and pseudo-label supervised self-distillation for rare disease imaging phenotype classification with dispersion-aware imbalance correction8
A novel hybrid approach to rare disease imaging phenotype classification is proposed, featuring three key novelties targeted at the above drawbacks, which substantially outperforms existing FSL methods via effective integration of the URL, pseudo-label driven self-distillation, and dispersion-aware imbalance correction, thus establishing a new state of the art.
- TAKT: Target-Aware Knowledge Transfer for Whole Slide Image Classification8
This paper proposes a Target-Aware Knowledge Transfer framework, employing a teacher-student paradigm, and introduces a target-aware feature alignment module to establish a transferable latent relationship between the source and target features by solving the optimal transport problem.
- Causally Debiased Time-aware Recommendation8
A debiased framework based on the inverse propensity score (IPS) and extended to the doubly robust method to accurately learn user preference is proposed and a sensitivity analysis method is developed to obtain more accurate IPS.
- 8
- 8
- Improving GAN Training via Feature Space Shrinkage8
This paper proposes a simple yet effective module, namely AdaptiveMix, for GANs, which shrinks the regions of training data in the image representation space of the discriminator, and proposes to construct hard samples and narrow down the feature distance between hard and easy samples.
- 8
- Medical Symptom Detection in Intelligent Pre-Consultation Using Bi-directional Hard-Negative Noise Contrastive Estimation8
This work formulate symptom detection as a retrieval problem and proposes a bi-directional hard-negative enforced noise contrastive estimation method (Bi-hardNCE) to tackle the symptom detection problem.
- Assessing cardiac function from total-variation-regularized 4D C-arm CT in the presence of angular undersampling8
This work examines how the quality of spatially and temporally total-variation-regularized reconstruction degrades when using as few as 6.9±1.2 projection views per heart phase and applies a convex optimization approach based on primal-dual splitting to 4D cardiac C-arm computed tomography.
- Model-Based 3D Cardiac Image Segmentation With Marginal Space Learning8
This chapter presents a model-based approach to extracting the major coronary arteries using the segmented cardiac chambers to predict the initial centerline path, and presents a hybrid coronary centerline extraction approach, robust under severe stenosis and flexible enough to handle anatomical variations.
- Model-Driven Centerline Extraction for Severely Occluded Major Coronary Arteries8
The proposed model-driven approach to extracting the three major coronary arteries can handle variations in the length and topology of an artery, and it is more robust under severe occlusions than a data- driven approach.
- Automatic Extraction of 3D Dynamic Left Ventricle Model from 2D Rotational Angiocardiogram8
An automatic method to directly extract 3D dynamic left ventricle (LV) model from sparse 2D rotational angiocardiogram (each cardiac phase contains only five projections) is proposed and provides quantitative cardiac function for analysis.
- MedREK: Retrieval-Based Editing for Medical LLMs with Key-Aware Prompts7
This work proposes MedREK, a retrieval-based editing framework that integrates a shared query-key module for precise matching with an attention-based prompt encoder for informative guidance and provides the first validated solution for batch-editing in medical LLMs.
- A Refreshed Similarity-Based Upsampler for Direct High-Ratio Feature Upsampling7
A refreshed similarity-based feature upsampling framework named ReSFU is constructed, which consistently achieves satisfactory performance on six tasks, including semantic segmentation, medical image segmentation, instance segmentation, panoptic segmentation, object detection, and monocular depth estimation, showing superior generality and ease of deployment beyond the existing upsamplers.
- 7
- 7
- 7
- 7
- ICA-Proto: Iterative Cross Alignment Prototypical Network for Incremental Few-Shot Relation Classification7
A novel model named ICA-Proto: Iterative Cross Alignment prototypical network is proposed, which incorporates the query representation into the encoding of novel prototypes and utilizes the query-aware prototypes to update the query represented at the same time.
- 7
- ClassFormer: Exploring Class-Aware Dependency with Transformer for Medical Image Segmentation7
A novel transformer, called ClassFormer, powered by two appealing transformers, i.e., intra-class dynamic transformer and inter-class interactive transformer, to address the challenge of fully exploration on compactness and discrepancy on medical image segmentation.
- Prompt Combines Paraphrase: Teaching Pre-trained Models to Understand Rare Biomedical Words7
Experimental results show that the proposed simple yet effective approach to helping models learn rare biomedical words during tuning with prompt can achieve up to 6% improvement in biomedical natural language inference task without any extra parameters or training steps using few-shot vanilla prompt settings.
- Lifelong Learning Based Disease Diagnosis on Clinical Notes7
Attention is proposed to adopt attention to combine medical entities and context, embedding episodic memory and consolidation to retain knowledge, such that the learned model is capable of adapting to sequential disease-diagnosis tasks.
- PRASEMap7
PRASEMap is presented, an unsupervised KG alignment system that iteratively computes the Mappings with both Probabilistic Reasoning (PR) And Semantic Embedding (SE) techniques.
- Automatic Heart Isolation in 3D CT Images7
An automatic heart segmentation algorithm for the diagnosis of coronary artery diseases (CAD) that combines a model-based detection framework with data-driven post-refinements to create a mask for a given cardiac CT image that contains only the relevant part of the heart.
- Aortic valve and ascending aortic root modeling from 3D and 3D+t CT7
This paper proposes a novel approach for morphological and functional quantification of both the aortic valve and the ascendingAortic root, estimated from CT, for the first time, to the best of the authors' knowledge.
- 6
- Improving Biomedical Entity Linking with Retrieval-Enhanced Learning6
A new scheme kNN-BioEL is introduced, which provides a BioEL model with the ability to reference similar instances from the entire training corpus as clues for prediction, thus improving the generalization capabilities.
- 6
- Tell Me How to Survey: Literature Review Made Simple with Automatic Reading Path Generation6
A new task named Reading Path Generation (RPG) is introduced which aims at automatically producing a path of papers to read for a given query and a graph-optimization-based approach for reading path generation which takes the relationship between papers into account is proposed.
- Training Automatic View Planner for Cardiac MR Imaging via Self-supervision by Spatial Relationship Between Views6
This work presents a clinic-compatible and annotation-free system for automatic CMR view planning that achieves superior accuracy to existing approaches including conventional atlas-based and newer deep learning based ones, in prescribing four standard CMR views.
- Integrating Semantics and Neighborhood Information with Graph-Driven Generative Models for Document Retrieval6
Under the approximation, it is proved that the training objective can be decomposed into terms involving only singleton or pairwise documents, enabling the model to be trained as efficiently as uncorrelated ones, and achieves superior performance over state-of-the-art methods.
- Alternative Baselines for Low-Shot 3D Medical Image Segmentation---An Atlas Perspective6
It is argued that low-shot segmentation on medical images should step further to explicitly learn dense correspondences between images to utilize the anatomical similarity and shed light on further understanding of atlas-based few- shot segmentation.
- 6
- Multimodal medical volumes translation and segmentation with generative adversarial network6
This work proposes an end-to-end 3D convolutional neural network (CNN) composed of mutually-beneficial generators and segmentors for image synthesis and segmentation tasks and shows both tasks are beneficial to each other and coupling these two tasks results in better performance than solving them exclusively.
- Interventional heart wall motion analysis with cardiac C-arm CT systems6
A complete framework to study the ventricular wall motion in 4D (3D+t) directly in the catheter lab and to discriminate regionally pathological from normal myocardium is presented.
- Patient-Specific Modeling of the Heart: Applications to Cardiovascular Disease Management6
A novel physiological model of the complete heart, including the chambers and valvular apparatus is introduced, which captures a large spectrum of morphological, dynamic and pathological variations and enables for patient-specific hemodynamic simulations and blood flow analysis.
- Towards the construction of a virtual yeast5
A virtual yeast is proposed, an artificial intelligence (AI)-driven agent that models eukaryotic cellular behaviours by integrating multimodal biological data, mechanistic reasoning and active experimentation using Saccharomyces cerevisiae as a genetically tractable and data-rich model system.
- CLIMS++: Cross Language Image Matching with Automatic Context Discovery for Weakly Supervised Semantic Segmentation5
A novel framework of Cross Language Image Matching with Automatic Context Discovery (CLIMS++), based on the recently introduced Contrastive Language-Image Pre-training (CLIP) model, for WSSS is proposed to introduce natural language supervision to activate more complete object regions and suppress class-related background regions in CAM.
- Guiding Large Language Models for Biomedical Entity Linking via Restrictive and Contrastive Decoding5
- Single Domain Generalization for Multimodal Cross-Cancer Prognosis via Dirac Rebalancer and Distribution Entanglement5
This work is the first to reveal that multimodal prognosis models often generalize worse than unimodal ones in cross-cancer scenarios, despite the critical need for such robustness in clinical practice.
- TopoTTA: Topology-Enhanced Test-Time Adaptation for Tubular Structure Segmentation5
Topology-enhanced Test-Time Adaptation (TopoTTA), the first test-time adaptation framework designed specifically for TSS, is proposed, which improves topological continuity by a novel Topology Hard sample Generation (TopoHG) strategy and prediction alignment on hard samples with pseudo-labels in the generated pseudobreaks.
- Deep-learning tool for early identification of non-traumatic intracranial hemorrhage etiology and application in clinical diagnostics based on computed tomography (CT) scans5
The proposed deep learning tool can be an accuracy tool for early identification of hemorrhage etiologies based on NCCT scans and may also provide more information for clinicians for triage and further imaging examination selection.
- 5
- FM-ABS: Promptable Foundation Model Drives Active Barely Supervised Learning for 3D Medical Image Segmentation5
This work proposes the Foundation Model-driven Active Barely Supervised (FM-ABS) learning paradigm for developing customized 3D specialist segmentation models with shoestring annotation budgets, i.e., merely labeling three slices per scan.
- GraphLeak: Patient Record Leakage through Gradients with Knowledge Graph5
This paper presents GraphLeak, which incorporates the medical knowledge graph in gradient leakage attacks, and has achieved state-of-the-art attack performance compared with previous works.
- 5
- Boost Supervised Pretraining for Visual Transfer Learning: Implications of Self-Supervised Contrastive Representation Learning5
This work empirically investigates when and why unsupervised pretraining surpasses supervised counterparts for image classification tasks with a series of control experiments and proposes Classification Activation Map guided contrastive (CAMtrast) learning which better utilizes the label supervsion to strengthen supervised pretraining.
- 5
- Alleviating Noisy-label Effects in Image Classification via Probability Transition Matrix5
A plugin module, namely noise ignoring block (NIB), consisting of a probability transition matrix and an inter-class correlation (IC) loss, to separate the hard samples from the mislabeled ones, and further boost the accuracy of image classification network trained with noisy labels.
- 5
- 5
- 5
- 5
- 4
- Empowering scientific discovery with explainable small domain-specific and large language models4
A forward-looking integration of explainable AI (XAI)-based research paradigms, encompassing small domain-specific models, large language models (LLMs), and agent-based large-small model collaboration is offered, highlighting the need for context-aware explanations tailored to research goals.
- 4
- 4
- NeRF2Points: Large-Scale Point Cloud Generation From Street Views' Radiance Field Optimization4
NeRF2Points adeptly navigates the inherent challenges of NeRF-based point cloud synthesis through the implementation of the following strategic innovations: Integration of Weighted Iterative Geometric Optimization (WIGO) and Structure from Motion (SfM) for enhanced camera pose accuracy, elevating street-view data precision.
- Generation is better than Modification: Combating High Class Homophily Variance in Graph Anomaly Detection4
A novel GNN model named HedGe, which improves the robustness under the novel Heterophily Attack with increased class homophily variance on other graph classification tasks, and improves the robustness under the novel Heterophily Attack with increased class homophily variance on other graph classification tasks.
- RECIST Weakly Supervised Lesion Segmentation via Label-Space Co-Training4
A pair of under- and over-segmenting masks are constructed for each lesion based on its RECIST annotation and served as the label for co- training a pair of subnets, respectively, along with the proposed label-space perturbation induced consistency loss to bridge the gap between the two subnets and enable effective co-training.
- Acquiring New Knowledge Without Losing Old Ones for Effective Continual Dialogue Policy Learning4
A novel continual dialogue-policy model is proposed which tackles the issues of “not forgetting the old” and “acquiring the new” from three different aspects: the forgetting preventor, the adaption accelerator, and a replay manager which helps regularly clean up the old data.
- 4
- 4
- Gait Cycle-Inspired Learning Strategy for Continuous Prediction of Knee Joint Trajectory from sEMG4
A model integrating two gait cycle-inspired learning strategies to mitigate the challenge for predicting human knee joint trajectory is proposed, which is the best performance in relevant literatures that has been reported, with reduced RMSE by at least 9.5%.
- A New Perspective to Boost Vision Transformer for Medical Image Classification4
The proposed Bootstrap Own Latent of Transformer (BOLT) is evaluated on three medical image processing tasks, and the experimental results validate the superiority of the BOLT for medical image classification, compared to ImageNet pretrained weights and state-of-the-art self-supervised learning approaches.
- 4
- 4
- Learning Shape Priors by Pairwise Comparison for Robust Semantic Segmentation4
An extra encoder is introduced into the classic encoder-decoder structure to form a Siamese structure for the encoders, to utilize both types of prior information encoded by the template-encoder and model the inter-subject similarity for each foreground class.
- Learning and Exploiting Interclass Visual Correlations for Medical Image Classification4
This paper presents the Class-Correlation Learning Network (CCL-Net) to learn interclass visual correlations from given training data, and produce soft labels to help with classification tasks via distance metric learning of class-specific embeddings with a lightweight plugin CCL block.
- Robust Landmark Detection in Volumetric Data with Efficient 3D Deep Learning4
An efficient and robust deep learning algorithm capable of full 3D detection in volumetric data is proposed and small 3D patches from a multi-resolution image pyramid are extracted to mitigate the over-fitting issue, thereby increasing detection robustness.
- Discriminative Learning for Anatomical Structure Detection and Segmentation4
This chapter summarizes several discriminative learning methods for detection and segmentation of anatomical structures and proposes innovative detector structures, namely Probabilistic Boosting Network and Marginal Space Learning, to address the challenges in anatomical structure detection.
- 3
- 3
- A plug and play fuzzy mask extraction module for single image deraining3
Comprehensive experiments show that combining the rain removal network with the proposed model not only improves the rain removal effect of the algorithm, but also gives clearer background details of the image.
- 3
- 3
- D-CAM: Learning Generalizable Weakly-Supervised Medical Image Segmentation from Domain-Invariant CAM3
- 3
- GM-ABS: Promptable Generalist Model Drives Active Barely Supervised Training in Specialist Model for 3D Medical Image Segmentation3
The Generalist Model-driven Active Barely Supervised (GM-ABS) learning paradigm is proposed, for developing specialized 3D segmentation models under extremely limited (barely) annotation budgets, e.g., merely cross-labeling three slices per selected scan.
- 3
- 3
- 3
- JoTR: A Joint Transformer and Reinforcement Learning Framework for Dialog Policy Learning3
This work introduces a novel framework, JoTR, which formulates a word-level policy that allows for a more dynamic and adaptable dialogue action generation, without the need for any action templates, and enhances the diversity of responses and improves the system's ability to handle edge cases effectively.
- 3
- 3
- Robust Representation via Dynamic Feature Aggregation3
With the regularization and orthogonal classifier, a more compact embedding space can be obtained, which accordingly improves the model robustness against adversarial attacks and can also achieve the state-of-the-art performance for out- of-distribution (OOD) detection, due to the learned compact feature space.
- Mix-and-Interpolate: A Training Strategy to Deal With Source-Biased Medical Data3
This work addresses an overlooked problem for training deep convolutional neural networks for COVID-19 classification using real-world multi-source data, namely, the data source bias problem and proposes MIx-aNd-Interpolate (MINI), a conceptually simple, easy-to-implement, efficient yet effective training strategy.
- Pericardium Based Model Fusion of CT and Non-contrasted C-arm CT for Visual Guidance in Cardiac Interventions3
A model based fusion method using the pericardium to align pre-operative CT to intra-operative C-arm CT is proposed, which can be applied to fuse different cardiac models (e.g., chambers, aorta, coronary arteries, and cardiac valves).
- Robust pigtail catheter tip detection in fluoroscopy3
A new tree-structured, hierarchical detection scheme, based on a shape categorization of the pigtail catheter tip, and a combination of novel Haar features is proposed, which demonstrates improved detection performance and validation on a data set consisting of 272 sequences with more than 20,000 images.
- 3D Lung Tumor Motion Model Extraction from 2D Projection Images of Mega-voltage Cone Beam CT via Optimal Graph Search3
A novel method to convert segmentation of objects with quasi-periodic motion in 2D rotational cone beam projection images into an optimal 3D multiple interrelated surface detection problem, which can be solved by a graph search framework is proposed.
- Handwriting Identification, Matching, and Indexing in Noisy Document Images3
This dissertation presents a rule line detection algorithm based on hidden Markov model (HMM) decoding, achieving both high detection accuracy and a low false alarm rate.
- 3
- 2
- Multimodal Mixture-of-Experts with Retrieval Augmentation for Protein Active Site Identification2
A reliability-aware fusion strategy based on Dempster–Shafer evidence theory that quantifies modality trustworthiness through belief mass functions and learnable discounting coefficients is proposed, enabling principled multimodal integration.
- Structure Observation Driven Image-Text Contrastive Learning for Computed Tomography Report Generation2
This work introduces a novel two-stage (structure- and report-learning) framework tailored for CTRG featuring effective structure-wise image-text contrasting and a dynamic, diversity-enhanced negative queue is proposed to guide the network in learning to discriminate various abnormalities.
- AtlantisGS: Underwater Sparse-View Scene Reconstruction via Gaussian Splatting2
AtlantisGS incorporates a scattering decomposition method that separates medium and object contributions during rendering, and a sparse Gaussian proliferation strategy that adaptively densifies the scene representation to improve structural accuracy.
- 2
- 2
- A Simple and Better Baseline for Visual Grounding2
A feature selection-based simple yet effective baseline for visual grounding, called FSVG, which directly encapsulate the linguistic and visual modalities into an overall network architecture without complicated iterative procedures, and introduces a similarity-based feature selection mechanism to only exploit language-related visual features for faster prediction.
- 2
- Beyond Single Models: Mitigating Multimodal Hallucinations via Adaptive Token Ensemble Decoding2
This work proposes Adaptive Token Ensemble Decoding (ATED), a training-free, token-level ensemble framework that mitigates hallucination by aggregating predictions from multiple LVLMs during inference and dynamically computes uncertainty-based weights for each model, reflecting their reliability at each decoding step.
- 2
- Unlocking the Potential of Weakly Labeled Data: A Co-Evolutionary Learning Framework for Abnormality Detection and Report Generation2
A co-evolutionary abnormality detection and report generation (CoE-DG) framework that utilizes both fully labeled and weakly labeled data to achieve mutual promotion between the abnormality detection and report generation tasks and introduces a bi-directional information interaction strategy with generator-guided information propagation (GIP) and detector-guided information propagation (DIP).
- 2
- 2
- Dynamically Masked Discriminator for Generative Adversarial Networks2
A novel method for GANs from the viewpoint of online continual learning is proposed, which automatically detects its retardation and then dynamically masks its features, such that the discriminator can adaptively learn the temporally-vary distribution of generated data.
- 2
- 2
- Decoupled Mixup for Generalized Visual Recognition2
This work proposes a novel "Decoupled-Mixup" method to train CNN models for OOD visual recognition, which enhances features from discriminative regions and suppress noise-prone ones when combining an image pair to improve the generalization ability of trained models.
- Generalized Brain Image Synthesis with Transferable Convolutional Sparse Coding Networks2
This paper presents a novel generalized brain image synthesis method, powered by the transferable convolutional sparse coding networks, to address the lack of interpretable cross-modal medical image representation learning.
- 2
- 2
- 2
- 2
- 2
- Deep convolutional neural networks for molecular subtyping of gliomas using magnetic resonance imaging2
Investigation of the use of deep convolutional neural networks for noninvasive glioma subtyping with radiological imaging data according to the new taxonomy announced by the World Health Organization in 2016 showed that the developed DCNN model can predict gliomas subtypes with promising performance, given sufficient, non-ill-balanced training data.
- Learning Crisp Edge Detector Using Logical Refinement Network2
A novel logical refinement network for crisp edge detection is proposed, which is motivated by the logical relationship between segmentation and edge maps and can be applied to both 2D and 3D images.
- 2
- 2
- 2
- Sparse Appearance Learning Based Automatic Coronary Sinus Segmentation in CTA2
This work proposes a multiscale sparse appearance learning based method for estimating vesselness towards automatically extracting the centerlines, which leverages the flexibility of sparse representation to model the inherent spatial coherence of vessel/background appearance and derive a vesselness measurement.
- Automatic 3D Motion Estimation of Left Ventricle from C-arm Rotational Angiocardiography Using a Prior Motion Model and Learning Based Boundary Detector2
A new learning-based boundary detector is developed by training a boosting boundary classifier combined with the principal component analysis of a local image patch to overcome ambiguities to LV endocardium boundary detection.
- Performance Evaluation of Spatial Domain Error Concealment for Image Recovery2
Evaluated results of spatial domain error concealment for image recovery show Maximally smoothness recovery approach has relatively good performance, and low computation complexity, and an adaptive version of MSR is proposed.
- 1
- IQE-CLIP: Instance-Aware Query Embedding for Zero-/Few-Shot Anomaly Detection in Medical Domain1
This work introduces class-based prompting tokens and learnable prompting tokens for better adaptation of CLIP to the medical domain, and designs an instance-aware query module (IQM) to extract region-level contextual information from both text prompts and visual features, enabling the generation of query embeddings that are more sensitive to anomalies.
- PRIME: A Pretrained Representation-Induced Model for 3D Molecules in De Novo Binder Design1
Semantics-Preserving Exploratory Sampling (SPES), which integrates Graph Laplacian Spectral Noise to respect chain connectivity and Conditional Freedom Modulation to dynamically balance exploration with fidelity, enables diversity-enhanced generation without sacrificing geometric validity under the reported structural metrics, improving the empirical exploration--fidelity trade-off.
- 1
- The Butterfly Effect in Pathology: Exploring Security in Pathology Foundation Models1
This work presents the first systematic investigation into the security of pathology foundation models for whole slide image~(WSI) analysis against adversarial attacks, and introduces the principle of local perturbation with global impact and proposes a label-free attack framework that operates without requiring access to downstream task labels.
- 1
- 1
- BrainSegDMIF: A Dynamic Fusion-enhanced SAM for Brain Lesion Segmentation1
A large-scale fully automated segmentation model specifically designed for brain lesion segmentation, named BrainSegDMIF, that has the following features: Dynamic Modal Interactive Fusion module that processes and integrates multi-modal data during the encoding process, providing the SAM encoder with more comprehensive modal information.
- 1
- 1
- Learning to Segment Multiple Organs from Multimodal Partially Labeled Datasets1
This work proposes a novel framework for learning to segment multiple organs from multimodal partially labeled datasets, which constructs a cross-modal a priori atlas from training data, which implicitly contains prior knowledge of organ locations, shapes, and sizes.
- Triplet-branch network with contrastive prior-knowledge embedding for disease grading1
A Triplet-Branch Network with ContRastive priOr-knoWledge embeddiNg (TBN-CROWN) for the accurate disease grading, which enables physicians to accordingly take appropriate treatments and yield a satisfactory grading accuracy for various diseases.
- 1
- URoadNet: Dual Sparse Attentive U-Net for Multiscale Road Network Extraction1
URoadNet offers a novel alternative to the U-Net architecture by integrating connectivity attention, which can exploit intra-road interactions across multi-level sampling features with reduced computational complexity.
- 1
- Deep-Learning Tool for Early Identifying Non-Traumatic Intracranial Hemorrhage Etiology based on CT Scan1
The proposed deep learning algorithms can be an effective tool for early identification of hemorrhage etiologies based on NCCT scans and may also provide more information for clinicians for triage and further imaging examination selection.
- 1
- Knowing the Past to Predict the Future: Reinforcement Virtual Learning1
This paper presents a cost-efficient framework, such that the RL model can evolve for itself in a Virtual Space using the predictive models with only historical data, and consistently outperforms the existing state-of-the-art methods.
- A Benchmark for Weakly Semi-Supervised Abnormality Localization in Chest X-Rays1
A regularization term, namely multi-point consistency, is proposed, which drives the model to generate the consistent bounding box from different point annotations inside the same abnormality, to deeply exploit the useful information from the weakly annotated data for abnormality localization.
- 1
- 1
- 1
- 1
- 1
- 1
- Optimal Mean Shape for Nonrigid Object Detection and Segmentation1
This chapter presents the computation of an optimal mean shape that can represent better the whole shape population and demonstrates the advantages of the approach on aortic valve landmark detection and whole-heart segmentation.
- Part-Based Object Detection and Segmentation1
This chapter demonstrates the robustness and accuracy of part-based object detection and segmentation on three applications, namely, left atrium segmentation in 3D C-arm CT, left ventricle detection in 2D MRI, and aorta segmentation
- Applications of Marginal Space Learning in Medical Imaging1
This chapter provides a review of marginal space learning applications in the published literature, first reviewing applications on “pure” detection problems, followed by those combining detection, segmentation, and tracking.
- Nonrigid Object Segmentation: Application to Four-Chamber Heart Segmentation1
This chapter presents an automatic object detection and segmentation framework based on Marginal Space Learning (MSL), which integrates the components described in previous chapters into a complete segmentation system.
- Constrained Marginal Space Learning1
In this chapter, the constrained MSL is presented to exploit correlations learned from the training set to increase the efficiency of the computational framework and employ the quaternion formulation for 3D orientation representation and distance measurement.
- Comparison of Marginal Space Learning and Full Space Learning in 2D1
A thorough comparison experiment on the LV detection in MRI images shows that the MSL significantly outperforms the FSL, in both speed and accuracy.
- 1
- 1
- Comorbidity-Aware Radiology Report Generation–
This work builds a hyperbolic contrastive language-image pre-training model to align image-text semantics; estimates class prototypes from single-label cases and performs prototype-driven hyperbolic contrastive learning to enforce radius ordering for multi-label samples; and incorporates a learnable hyperbolic soft reward into the group relative policy optimization framework to directly optimize clinical accuracy.
- –
- Prior-driven multimodal contrastive learning with Fourier cross-attention to detect aortic dissection in 3D non-contrast CT–
An anatomy prior-driven multimodal contrastive learning framework, extracting discriminative features from contrast-enhanced CT (CE-CT) and radiological reports to enhance NC-CT-based AD detection and enable accurate true/false lumen localization using NC-CT alone is developed.
- siProGenA: Generative siRNA Candidate Construction via Position Proposal and Guide Generation–
A modular proposal–generation view of siRNA candidate construction for prioritizing compact candidate sets is supported, using a Discrete Denoising Diffusion Probabilistic Model for mRNA-conditioned position proposal and a Bayesian Flow Network for temperature-controlled guide generation.
- Structure-guided prior-driven implicit neural representations for time-of-flight PET image reconstruction–
The proposed approach integrates INRs with physics-consistent modeling and prior-guided regularization, providing an effective unsupervised framework for TOF-PET reconstruction and highlighting the potential of neural field representations for tomographic imaging.
- An Effective Semi-Subject-Independent sEMG-Based Learning Framework to Continuously Predict Knee Joint Trajectory–
A semi-subject-independent deep learning framework that is pretrained on source subjects to learn shared cross-subject representations and then calibrated with only a few trials from an unseen target subject before testing on that subject's held-out trials is proposed.
- –
- –
- –
- X-ray Insights Unleashed: Pioneering the Enhancement of Multi-Label Long-Tail Data–
A novel data synthesis pipeline designed to augment tail lesions utilizing a copious supply of conventional normal X-rays, and the integration of a Large Language Model Knowledge Guidance module alongside a Progressive Incremental Learning strategy to stabilize the inpainting fine-tuning process is proposed.
- –
- Robust Federated Video-based Remote Physiological Measurement for Heterogeneous Multi-source Data–
This work characterize the heterogeneity of multi-source rPPG data from both the input and output domain perspectives and introduces pseudo-labeling techniques and proposes a novel FL framework, FedGRC, which incorporates automatic gradient regularization calibration to mitigate the impact of input domain discrepancies on model performance.
- –
- Robust Source-Free Domain Adaptation for Medical Image Segmentation based on Curriculum Learning–
A curriculum-based framework, namely learning from curriculum (LFC), for source-free domain adaptation, which consists of easy-to-hard and source-to-target curricula is proposed, which surpasses the existing approaches and achieves a new state-of-the-art.
- –
- –
- BrainSegDMlF: A Dynamic Fusion-enhanced SAM for Brain Lesion Segmentation–
A large-scale fully automated segmentation model specifically designed for brain lesion segmentation, named BrainSegDMLF, which has the following features: Dynamic Modal Interactive Fusion module that processes and integrates multi-modal data during the encoding process, providing the SAM encoder with more comprehensive modal information.
- –
- –
- –
- K-Space-Aware Cross-Modality Score for Synthesized Neuroimage Quality Assessment–
K-CROSS uses a pre-trained multi-modality segmentation network to predict the lesion location, together with a tumor encoder for representing features, such as texture details and brightness intensities, and outperforms other metrics, especially in comparison with the radiologists on NIRPS.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- IJCARS MICCAI 2015 Special Issue–
This special issue of the International Journal of Computer Assisted Radiology and Surgery (IJCARS) includes six research papers nominated from the papers presented at MICCAI 2015 in Munich, Germany, for which the review and revision process could be finalized in time for inclusion in this special issue.
- –
- –
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-14. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.