ECCV2026 Medical Imaging AI paper notes paper summaries Segmentation Diffusion Models Multimodal/VLM Super-Resolution Adversarial Robustness

🏥 Medical Imaging¶

🎞️ ECCV2026 · 110 paper notes

📌 Same area in other venues: 🧠 NeurIPS2026 (1) · 📷 CVPR2026 (173) · 🔬 ICLR2026 (87) · 🧪 ICML2026 (28) · 🤖 AAAI2026 (76) · 🧠 NeurIPS2025 (77)

🔥 Top topics: Medical Imaging ×55 · Segmentation ×21 · Diffusion Models ×10 · Multimodal/VLM ×8 · Super-Resolution ×5

Comprehensive language–image pre-training for 3D medical image understanding: COLIPRI combines chest CT image–text alignment, affirmative/negative short-statement discrimination, report generation, and multiresolution masked reconstruction so that one visual encoder captures both global semantics and local structure; the full model reaches 73.23% zero-shot AUROC on RAD-ChestCT, although not every task requires the report-generation branch.
MedSynapse-V: Bridging Visual Perception and Clinical Intuition via Latent Memory Evolution: MedSynapse-V compresses anatomical priors from a medical segmentation encoder into continuous memory, trains the model to use that memory through region-masking counterfactual rewards, and distills it into an autonomous branch without the external encoder, achieving a reported seven-benchmark average of 59.6 on offline medical evaluations, not evidence of clinical effectiveness.
FlexiBrain: Resolution-Agnostic Voxel-Level Encoding for Native fMRI: FlexiBrain defines fMRI patches in millimeters and seconds, combines dynamically resized shared kernels with physical positional encoding and Mamba-JEPA, and learns directly from heterogeneous native-space data, achieving the highest accuracy on four of five disease classification tasks while avoiding expensive template registration.
MOOZY: A Patient-First Foundation Model for Computational Pathology: MOOZY learns spatial slide representations through masked self-distillation, then jointly trains patient-level representations with a case transformer and 333 clinical tasks, achieving macro-average weighted F1 of 0.769 and balanced accuracy of 0.702 across 16 held-out tasks, while its macro-average AUC remains slightly below TITAN.
Render-FM: Feedforward Model for Real-time Photorealistic Volumetric Rendering: Render-FM proposes a feedforward model based on an nnU-Net-style 3D U-Net that directly regresses 6D Gaussian Splatting (6DGS) parameters from CT volume data in a single 2.8-second forward pass. This eliminates the per-scan optimization bottleneck (~500x speedup). By introducing Anatomy-Guided Prior Initialization (AGP), it utilizes segmentation masks and transfer functions as structural and appearance priors, bridging the domain gap between natural scene reconstruction and medical volume rendering. It supports real-time interactive rendering (328+ FPS) and organ-composition visualization with zero extra overhead.
ECHO: Efficient Chest X-ray Report Generation with One-step Block Diffusion: ECHO converts a medical autoregressive model into a block diffusion model and distills conditional distributions from multi-step teacher trajectories into a one-step-per-block student, retaining strong clinical-content scores on the paper's normalized chest X-ray report test sets while reaching 274.21 tokens/s with block size 8.
RAU: Reference-based Anatomical Understanding with Vision-Language Models: RAU trains Qwen2.5-VL-7B to compare anatomical spatial relationships between reference and target images, then connects segmentation tokens to SAM2 reference memory as soft prompts, achieving Dice scores of 0.7503 on CAMUS and 0.7010 on LERA when segmentation training uses only RAOS and Arcade.
Atlas is Your Perfect Context: One-Shot Customization for Generalizable Foundational Medical Image Segmentation: AtlasSegFM converts one annotated anatomical atlas into automatic prompts for a frozen segmentation foundation model and learns voxel-wise fusion at test time, raising SegRap Dice from 50.02 for five-click nnInteractive to 69.19 while retaining a dependence on anatomical correspondence between support and query.
Discrete Diffusion Models with MLLMs for Unified Medical Multimodal Generation: MeDiM adapts pretrained Liquid into a discrete diffusion model with bidirectional attention and timestep conditioning, using one denoising network for medical report-to-image, image-to-report, and joint image-report generation, with image-generation FID scores of 16.60 on MIMIC-CXR and 24.19 on PathGen.
MedSPOT: A Workflow-Aware Sequential Grounding Benchmark for Clinical GUI: MedSPOT is the first workflow-aware sequential grounding benchmark for clinical software interfaces: it records 216 interactive tasks with 597 annotated keyframes across 10 real medical imaging platforms, recasts GUI grounding as a 2-3 step state-dependent sequence of click decisions, and evaluates 16 MLLMs under a strict first-error-terminates protocol with a six-class failure taxonomy — the best model, GUI-Actor, reaches only 43.5% task completion accuracy, while human annotators reach 82.22%.
A Mechanism-Driven Theory of Phase Transitions in Active Learning: This paper proposes a mechanism-driven theory of phase transitions in active learning: the four components of the PAC generalization bound (empirical risk, distribution discrepancy, model complexity, confidence term) are reinterpreted as dynamically interacting terms, proving that a shift in their dominance along the labeling trajectory is inevitable. Measurable proxy metrics and piecewise regression are used to identify a "data-driven - transition - model-driven" three-phase taxonomy, explaining why representativeness, coverage, and uncertainty strategies excel at different budget stages.
BrainRiem: Riemannian Prototype Learning for Source-Free Cross-Site Brain Network Diagnosis: BrainRiem proposes to learn compact brain network prototypes on the Riemannian manifold via bi-level optimization, achieving cross-site functional brain connectivity diagnosis without access to source data. It substantially outperforms existing methods on the ABIDE and REST-meta-MDD multi-site benchmarks, and the learned prototypes exhibit biological interpretability.
FaCT-GS: Fast and Scalable CT Reconstruction with Gaussian Splatting: Rather than introducing another reconstruction network, FaCT-GS redesigns initialization and GPU computation for Gaussian CT reconstruction, reducing R2-Gaussian's 9 minutes to 1 minute 52 seconds while improving SSIM from 0.946 to 0.950 in the 75-view setting of Table 1.
Histopathology Multi-modal Embedding for Pathology Composed Retrieval: HOMIE proposes a model-agnostic two-stage adaptation framework to transform any generative MLLM into a pathology retrieval expert. In the first stage, pure-text LoRA training is employed to teach the LLM a discriminative metric space (resolving task mismatch). In the second stage, native resolution processing, stain augmentation, and a progressive knowledge curriculum inject pathomorphological priors (resolving domain mismatch). On the newly proposed PCR composed retrieval benchmark, a 2B parameter version significantly outperforms 7B specialized pathology MLLMs and dual-encoder models.
Low-Level Dataset Distillation for Medical Image Enhancement: The paper proposes the first dataset distillation framework for low-level medical image enhancement: a shared anatomical prior built from one representative subject is personalized per subject by a learnable subject-specific code (SPG), and subject-aware gradient matching aligns training trajectories, so that distilled data at roughly 99% compression approaches full-data training on CT/MRI super-resolution and low-dose CT restoration.
MediRound: Multi-Round Entity-Level Reasoning Segmentation in Medical Images: MediRound trains a medical dialogue model on MR-MedSeg to reference previously segmented entities, then improves masks passed across rounds with the Judgment & Correction Mechanism (JCM), increasing overall validation cIoU from 55.8 to 58.9.
Seeing What Matters: Lesion-Aware High-Resolution Patch Discovery and Fusion for Chest X-ray Report Generation: LePaX casts high-resolution perception for chest X-ray report generation as a spatial resolution allocation problem under a fixed visual token budget: a lightweight policy head predicts a lesion utility map over the global feature grid and selects Top-K 384×384 native-resolution patches, whose features are written back into the global grid at their original coordinates (no token growth), setting new NLG and clinical results on MIMIC-CXR / IU-Xray / CheXpertPlus while using over 10× fewer visual tokens than naive high-resolution tiling.
VesselTok: Tokenizing Vessel-like 3D Biomedical Graph Representations for Reconstruction and Generation: VesselTok dilates a 3D centerline graph of vessels or airways into a continuous "graph occupancy field" with a fixed pseudo-radius, then learns fixed-length compact tokens (512 × 4) with a VAE-style transformer over centerline points, so that a single latent space supports high-fidelity reconstruction, cross-anatomy generalization, vascular graph generation, and missing-link infilling.
3D Field of Junctions: A Noise-Robust, Training-Free Structural Prior for Volumetric Inverse Problems: This paper proposes the 3D Field of Junctions (3D FoJ), generalizing the explicit geometric parameterization of 2D FoJ (ICCV 2021) to three-dimensional volumes. By modeling each 3D patch as a junction where three planes intersect at a vertex, it performs patch-wise initialization via discrete coordinate descent followed by global joint gradient refinement. Without requiring any training data, it preserves sharp edges and corners under extremely low SNR conditions. It can also be embedded as a plug-and-play proximal regularization term into arbitrary volumetric inverse problems (validated on low-dose CT, cryo-ET, and point cloud denoising).
AlphaRad: Grounded Zero-Shot Classification in Chest Radiology via α-Corrected Binary Cross Entropy and Factorized Latent Supervision: AlphaRad turns chest radiology reports into per-image concept supervision and combines negative-sampled α-corrected binary cross entropy with independently supervised latent subspaces, reaching 83.3% average AUROC across 16 classification datasets at 224 resolution while improving spatial grounding.
Beyond Random Sampling: Distribution-Aware Alignment for Semi-Supervised Medical Image Segmentation: The framework selects annotation candidates using frozen visual features and density-weighted K-Center, then trains with memory-guided Copy-Paste and progressive branch activation, reaching 87.3% DSC on PROMISE with two labeled cases; the gain combines changed sample selection and changed training.
BioMedVR: Confusion-Aware Mixture-of-Prompt Experts for Biomedical Visual Reprogramming: BioMedVR introduces visual reprogramming (VR) to the biomedical imaging field for the first time. Through a dual-expert decoupled confusion-aware MoPE architecture—where a positive expert handles main class discrimination and a negative expert suppresses easily confused classes with the help of LLM-generated confusion attributes, coupled with a margin-based confusion suppression loss (CS Loss) to explicitly widen the gap between positive and negative scores—it consistently outperforms existing VR and prompt learning methods across 11 medical datasets (spanning 9 imaging modalities) and 7 natural image benchmarks, using only 1/500 of the parameters of full-model fine-tuning.
BrainFIBRE: A Foundation Model via Information Decomposition for Brain Microstructure: BrainFIBRE is the first foundation model for brain tissue microstructure. It treats three NODDI-derived microstructural maps (NDI/ODI/FWF) as three modalities. Using "Self-supervised Partial Information Decomposition (SPID) + Counterfactual Candidate Construction (CCC)", it pre-trains a Mixture-of-Experts (MoE) network on 55,000 subjects from the UK Biobank. It decouples the unique, redundant, and synergistic information of the three modalities without labels, achieving state-of-the-art (SOTA) performance across diverse downstream tasks including age, sex, cerebrovascular and neurodegenerative markers, and cognition in multi-ethnic cohorts. In addition, the expert weights are highly interpretable.
CAR-MIL: Counterfactual Attention Regularization for Multiple Instance Learning: To address the long-standing challenges of unfaithful attention attribution and lack of contradictory evidence in Multiple Instance Learning (MIL), CAR-MIL introduces a dual-branch counterfactual attention regularization framework that enforces prediction divergence under minimal, structured attention perturbations, decoupling complementary supporting and refuting evidence distributions while boosting classification performance across five digital pathology benchmarks and synthetic tasks.
CrossFeat: Bridging Imaging Modalities in Feature Descriptor Space: CrossFeat introduces a lightweight framework to learn a non-linear crossing operator in feature descriptor space by disentangling monomodal descriptors into an invariant geometry code and a modality-specific variational appearance code, enabling off-the-shelf descriptors to perform cross-modal matching across medical, driving, and satellite domains without retraining.
Decoding Children’s Gait Behavior: To overcome the subjectivity of pediatric neuromuscular gait screening and the inability of existing adult-centric foundation models to resolve subtle motor anomalies, this paper introduces the first open-source pediatric gait video dataset CGV (110 subjects, 1,185 multi-view videos with bilateral EVGS items) and develops ChildGait-Video, an end-to-end video architecture incorporating token-level kinematic prompts and mask-guided patch pruning that reaches an average accuracy of 84% across all 34 clinical scoring items.
DRIFT: Difficulty-aware Rectified Flows for Through-plane MRI Super-Resolution: To tackle stair-step artifacts and continuous slice-thickness degradation in anisotropic MRI, DRIFT introduces a two-stage framework combining an Anatomical Projection Network with residual rectified flow, physics-aware difficulty (PAD) step scheduling, and consistent endpoint trajectory alignment (CETA), delivering high-fidelity isotropic reconstructions with substantially reduced function evaluations.
Dual-Prior Guided Null-Space Learning with Mixture-of-Splines for Arbitrary Medical Slice Super-Resolution: This paper reformulates arbitrary-scale medical slice super-resolution as a constrained inverse problem. By utilizing range-null space decomposition, it achieves a hard constraint for measurement consistency. Within the null space, a Mixture-of-Splines (MoS) is employed to adaptively allocate analytical continuity orders according to local anatomical features, strictly guaranteeing that the acquired slices remain unchanged while generating anatomically plausible interpolation details.
Equivariant Symmetry-Aware Head Pose Estimation for Fetal MRI: E(3)-Pose explicitly embeds E(3)-equivariant convolutional filters and models the bilateral anatomical symmetry axis via an even-parity pseudovector, achieving robust 6-DoF fetal head pose estimation from highly artifact-corrupted clinical 3D MRI navigators in 0.3s for real-time adaptive slice prescription.
From Multi-Resolution Cells to Gigapixel Whole Slide Images Foundation Model for Computational Pathology: Addressing the limitation of existing computational pathology foundation models that operate predominantly on single-resolution inputs and lack hierarchical multi-resolution modeling, this paper introduces the Multi-Resolution Pyramid Transformer (MRPT). By combining a lightweight Consecutive Cross-Resolution Attention mechanism with a three-stage self-supervised learning paradigm across cellular, patch, regional, and whole-slide hierarchies, MRPT and its multimodal derivative MRPT-LLaVA establish substantial new state-of-the-art results across 34 clinical benchmarks.
From Reconstruction to Decision: A Post-Encoder Plug-in Adapter for Curvilinear Segmentation: This paper proposes PEPA, a lightweight post-encoder plug-and-play adapter. Through two modules—Target-Conditioned Serpentine Upsampling (TCSU) and Target-Adaptive Differentiable Binarization (TADT)—it significantly improves the topological continuity of curvilinear structure segmentation (such as blood vessels and cracks) while keeping the foundation model encoder frozen. With an increase of only 0.26M parameters, it achieves average improvements of +2.6% IoU and +2.8% clDice across five medical and industrial benchmarks.
Gaussian Volumetric Representation for Efficient Shear–Warp Visualization: To overcome the prohibitive cost of ray marching and the surface limitation of standard Gaussian splatting on dense medical volumes (MRI and Visible Korean Cryosection), this paper presents a continuous Gaussian volumetric representation optimized via unbiased Monte Carlo importance sampling and slice-based curriculum learning, coupled with an adaptive single-stack GPU shear–warp renderer that achieves up to 43.86 FPS rendering and an 11.31:1 compression ratio.
IConE: Batch Independent Collapse Prevention for Self-Supervised Representation Learning: IConE decouples the anti-collapse mechanism in self-supervised learning from transient mini-batch statistics to a global learnable instance embedding table regularized by an explicit diversity objective, enabling stable pretraining down to a batch size of 1 with superior robustness against severe class imbalance.
IP-SAM: Rethinking Prompt-Conditioned Segmentation for Prompt-Absent Deployment: Addressing the structural interface mismatch where foundation segmenters lack external prompts during automatic deployment, IP-SAM proposes a prompt-space conditioning paradigm that synthesizes complementary intrinsic foreground/background prompts through a frozen prompt encoder and applies asymmetric prompt-space gating to suppress background false positives, achieving state-of-the-art performance with only 21.26M trainable parameters.
Linguistically-Aligned and Visually-Grounded Preference Optimization for Clinically-Augmented Medical Report Generation: Addressing the issues of conventional Direct Preference Optimization (DPO) in medical report generation—namely the confounding of clinical facts with stylistic variations and the absence of visual grounding—this paper proposes DPO-Clin, which utilizes entity-level clinical diagnostics to craft linguistically-aligned report pairs, introduces context-triggered preference inversion (M2DPO), and suppresses latent clinical risks through counterfactual substitution.
MedGSSR: Generalizable Medical Image Super-Resolution 3D Reconstruction via Hierarchical Feed-forward Gaussian Splatting: Addressing the core limitations of per-subject optimization, spectral bias in coordinate-based neural implicit representations, and slice-wise inconsistencies of 2D proxy supervision, MedGSSR introduces a generalized, purely volumetric feed-forward 3D Gaussian Splatting framework that factorizes anatomy into coarse structure and fine texture via sub-voxel decomposition, enabling fast, test-time optimization-free, arbitrary-scale 3D super-resolution with exceptional cross-dataset generalizability.
Memory-Supported Synergistic Adaptation for Training-Free Test-Time Medical Image Segmentation: Addressing the vulnerability of test-time fine-tuning in vision-language models where noisy updates degrade pretrained representations, this paper proposes MSSA, a training-free synergistic adaptation framework that stabilizes coarse predictions, curates a noise-aware dynamic memory bank, and conducts cross-image prototype matching for structural guidance.
Mind2Cloud: EEG-to-Point Cloud Generation with Two-Granularity Diffusion Decoding: Addressing the evolving semantic discrepancy between brain neural representations and diffusion denoising trajectories, Mind2Cloud introduces a two-granularity point cloud decoder coupling a global Transformer with a local PVCNN branch via a timestep-adaptive fusion mask and adversarial refinement, substantially outperforming prior art in geometric fidelity and semantic alignment at 2048-point resolution across all 12 subjects.
Multi-Channel Uncertainty-Weighted Score Matching for Conditional Diffusion in Medical UDA: To overcome severe pseudo-label noise that destabilizes conditional diffusion models in medical unsupervised domain adaptation, this paper proposes an uncertainty-reweighted multi-channel score matching objective supported by theoretical minimum-MSE convex combination guarantees and feature-guided Bézier intensity alignment, synthesizing high-fidelity target-style labeled pairs that substantially boost cross-modality segmentation.
NGPS: Structure-Preserving Self-Supervised Denoising via Neighbor-Guided Patch Sampling: NGPS proposes a lightweight neighbor-slice self-supervised denoising framework: for inter-slice misaligned regions, local patch matching is performed on a noise-attenuated guidance map, while the training target retrieves pixel values from the corresponding matched coordinates in the original noisy neighbor-slices. This decouples "structural matching" from "signal retrieval" without requiring dense deformation field estimation, consistently improving fidelity and structure-sensitive metrics on low-dose CT and MRI.
OBBSeg: Irregular Lesion Segmentation under Oriented Bounding Box Annotations: To tackle the lack of orientation constraints and severe rectangular shape bias in traditional weakly supervised medical image segmentation, OBBSeg introduces an oriented bounding box (OBB) supervision framework equipped with a differentiable Mask-to-OBB (M2O) loss that eliminates geometric contour bias, alongside encoder-level prompt-assisted foreground enhancement and differential contrast modules, delivering segmentation accuracy on par with fully supervised methods at minimal annotation cost.
Physics-Grounded Disentangled Flow Modeling for Brain Disease Progression Trajectory: Addressing the physical implausibility caused by entangling structural deformation with image intensity changes in longitudinal brain disease forecasting, this paper proposes PDF, a disentangled flow matching framework regularized by Fisher-KPP PDE subspace projection, achieving state-of-the-art lesion boundary accuracy and trajectory consistency across three glioma and multiple sclerosis datasets.
Posterior Samplings are Missing Modalities Generators for Medical Image Translation: Formulating missing modality generation in multimodal medical imaging as a linear inverse problem under a joint distribution, this work trains an unconditional joint flow matching prior and applies measurement-consistent posterior guidance along with a many-to-one sampling schedule to synthesize arbitrary missing modalities with high anatomical fidelity without task-specific retraining.
Progression as Latent Drift: Generative Forecasting of Slow-Evolving Pathologies: To overcome the dominance of stationary anatomy and dense imaging noise in longitudinal brain MRI forecasting, Latent Drift reformulates generation to target temporal latent residuals combined with Finite Scalar Quantization (FSQ) acting as a topological dead-zone filter, enabling high-fidelity and clinically aligned individualized neurodegeneration forecasting.
Provable and Robust Wavefront Sensing via Self-Reference Interferometry: This paper introduces an analytical, non-iterative wavefront sensing framework based on spatially shifted self-reference interferometry, proving that consecutive co-prime shifts create a connected measurement graph with minimal hop diameter to bound worst-case error accumulation, and combining hop-constrained path averaging with closed-form Fourier least-squares refinement to achieve fast, provably robust phase recovery from as few as 8 to 16 measurements.
REVA-PO: Stabilizing Reinforcement Learning for Chest X-ray Report Generation: To resolve restricted exploration and policy collapse in reinforcement learning for chest X-ray report generation, REVA-PO introduces response-weighted KL regularization and periodic validation-anchored policy resets on a pretrained LVLM, establishing new state-of-the-art benchmarks in both clinical factual accuracy and linguistic quality.
SAM+D: Parameter-Efficient Dimensional Lifting of SAM-Family Models via Depth-Routed LoRA and Depth Shifting: SAM+D introduces a unified parameter-efficient dimensional lifting framework that inserts Depth-Routed LoRA (DRLoRA) experts and zero-parameter Depth Shift Modules (DSM) into frozen transformer blocks, tuning only ~2.8% to 3.7% of parameters to lift 2D SAM to 3D volumetric CT segmentation and 2D+T SAM2 to 4D spatiotemporal cell tracking.
Semantic-Anchored Evidential Fusion for Domain-Robust Whole-Slide Survival Analysis: To tackle catastrophic performance degradation in whole-slide image (WSI) survival analysis caused by cross-center staining and scanner domain shifts, SAEFS extracts domain-invariant semantic anchors via template-structured VQA to guide visual feature extraction and fuses multimodal evidence via Dirichlet subjective logic and a cautious conjunction rule, achieving state-of-the-art zero-shot cross-center prognosis and reliable calibration.
Spectral Consistent Flow for One-step 3D Medical Image Translation: Presents Spectral Consistent Flow (SC-Flow), an efficient 1-NFE 3D medical image translation framework that combines a stochastic Brownian bridge with a Spectral Consistency Corrector (SCC) to resolve high-frequency attenuation, over-smoothing, and modality entanglement in latent average velocity fields.
SpEmoC: A Balanced Speaker-Segment Multimodal Emotion Benchmark: Addressing the severe class imbalance and actor/scene leakage in existing conversational multimodal emotion benchmarks, SpEmoC introduces a large-scale benchmark of 30,000 speaker-segment clips across seven balanced emotion categories from 3,100 movies with strict movie-level partitioning, delivering robust cross-dataset and minority-emotion generalization.
TaxoMIL: Taxonomy-Constrained Learning for Hierarchical Whole Slide Image Analysis: TaxoMIL reformulates whole slide pathological image diagnosis as a hierarchical text generation task constrained by clinical taxonomies. Using a dual-head decoder, it simultaneously generates coarse- and fine-grained diagnostic texts. It enforces alignment of visual and label representation spaces with the medical taxonomy tree via hierarchical and image-text alignment losses, comprehensively outperforming existing methods on three WSI datasets: gastric, breast, and prostate.
TopoAgent: An Agentic Framework for Automated Topology Learning in Medical Imaging: Addressing the trial-and-error bottleneck in selecting and configuring topological descriptors for medical images, TopoAgent introduces the first LLM-based agentic framework operating on a Perception–Reasoning–Action–Reflection loop with distilled domain skills, achieving 68.21% average balanced accuracy across 26 datasets in TopoBenchmark.
Two-Parameter Flow Map Learning for Continuous-Time Diffeomorphic Image Registration: Addressing the reliance of non-autonomous diffeomorphic registration on explicit numerical integration and heavy time discretization, TPFM-DIR proposes directly learning the continuous-time solution operator of non-autonomous ODEs as a two-parameter flow map; by regularizing with anchored cocycle consistency, it eliminates velocity integration during training and achieves high-accuracy, topology-preserving alignment via minimal progressive compositions at inference.
Unified Multi-Layer Subspace Modeling for Cross-Domain OOD Detection: PRISM is a tuning-free post-hoc OOD detection method that discards brittle per-layer score aggregation and OOD-calibrated weighting in favor of a "fuse-then-model" principle, constructing a unified shallow-to-deep feature embedding whose ID principal subspace provides both projected class-conditional Mahalanobis distances and orthogonal residual energy scores, achieving a state-of-the-art 93.92% cross-domain average AUROC across natural, medical, and industrial vision benchmarks.
Unsupervised Source-Free Ranking of Biomedical Segmentation Models Under Distribution Shift: Addressing the acute shortage of biomedical annotations and pronounced domain shifts, this paper introduces CMR, the first black-box, unsupervised, source-free ranking framework for semantic and instance segmentation models, achieving high correlation with ground-truth rankings via only two inference passes measuring prediction consistency under perturbations.
XDen-1K: A Density Field Dataset of Real-World Objects: XDen-1K combines biplanar X-ray scans with part-level differentiable optimization to reconstruct volumetric density fields for 1,000 real-world objects (137 categories), paired with watertight meshes, part-level annotations, and real X-ray images, and defines three downstream tasks — a density-estimation benchmark, X-ray-conditioned volumetric segmentation, and center-of-mass-aware grasping, where adding its CoM prior raises AnyGrasp's success rate from 27.50% to 77.75%; reconstruction accuracy, however, is currently supported only by reprojection consistency and a small CT subset used as a partial reference.
XPos3R: Cross-Modal Transformer for Intraoperative 2D/3D Registration: XPos3R extends the DUSt3R-style feedforward geometry transformer to heterogeneous "2D X-ray + 3D CT" inputs: a pretrained RayDINO encodes the image, a 3D CNN trained from scratch encodes the volume, and a unidirectional image-to-volume cross-attention decoder directly regresses the intraoperative camera pose. A single pelvic pretraining works across patients, and paired with seconds of test-time optimization it reduces 3D error to about 3.3 mm and reprojection error to 0.45 mm, eliminating hours of per-patient preoperative retraining.
A Dual-space Patch-driven Complementary Learning Framework for Semi-supervised Multi-organ Segmentation: DPCL combines rarity- and uncertainty-guided cropping, full-image–patch pseudo-label fusion, and complementary contrastive objectives to reduce confusion between neighboring organs, reaching 68.89% mean Dice on Synapse with two labeled training cases.
Active Viewpoint Selection for Tomography via Perturbed Gaussian Ensemble: To address the active viewpoint selection challenge for X-ray in sparse CT, this paper proposes a perturbed Gaussian ensemble framework based on radiative 3DGS. By randomly perturbing the density parameters of low-density Gaussian primitives, a lightweight model ensemble is constructed. Epistemic uncertainty is quantified using the structural similarity (SSIM) variance of predictions under candidate viewpoints, thereby selecting the viewpoint that best exposes geometric artifacts as the next optimal acquisition angle.
Beyond Isolated Scans: Cross-Phase Alignment of Structure and Topology for 3D Medical Pretraining: CAST adds wavelet-based structural alignment and local feature-relation alignment to reconstruction-based pretraining on paired non-contrast and contrast-enhanced CT, reaching 0.9163 Dice on AIIB23 airway segmentation without outperforming every baseline on every downstream task.
CerDETR: Cell-Prior Empowered DETR for Cervical Lesion Detection: CerDETR adds a cell-prior correction branch during DINO training, combining multiscale proposals, overlap-and-containment matching, and category-scale query enhancement to improve cervical lesion detection from 30.9 to 39.1 AP on CDetector, while retaining only the main detection branch at test time.
CLARITY: Medical World Model for Guiding Treatment Decisions by Simulating Context-Aware Disease Trajectories in Latent Space: CLARITY presents a medical world model for longitudinal cancer treatment planning that simulates counterfactual disease trajectories directly in a compact 3D MRI latent space conditioned on irregular time intervals and clinical context, coupled with an entropy-regularized inverse survival feedback loop to iteratively optimize sequential therapy regimens.
CortexVideo: A Semantic-Spatial Dual-Anchor Framework for High-Fidelity fMRI-to-Video Reconstruction: Inspired by the visual dual-stream hypothesis, CortexVideo establishes a semantic-spatial dual-anchor fMRI-to-video reconstruction framework that leverages ventral-stream adaptive semantic decoding to prevent keyframe caption drift and dorsal-stream continuous soft heatmap regularization to enforce fine-grained subject localization, achieving state-of-the-art semantic fidelity and spatial accuracy on the cc2017 benchmark.
CS-TTA: Preserving Concept Sensitivity in Test-Time Adaptation: Addressing the hidden danger that unsupervised test-time adaptation (TTA) can exacerbate reliance on spurious shortcuts while improving overall accuracy, this paper formalizes Concept Sensitivity Drift (CSD) via TCAV directional derivatives and proposes CS-TTA—a plug-in regularizer that penalizes excessive sensitivity drift to anchor feature representations to task-relevant concepts without requiring target labels or architectural changes.
Decoding Multimodal Causality: End-to-End Multimodal Mediation Pathways Inference: Rather than fixing one causal graph, MM-SIGMA learns an ensemble in multimodal latent space, selects stable mediation pathways, and aggregates effects and variances, achieving 0.74 pathway recovery and 0.94 coverage for nominal 95% intervals in the paper's end-to-end synthetic evaluation; its clinical pathways remain observational findings contingent on identification assumptions.
DiffuPrompt: Adapting Video Foundation Models to 3D Medical Volumes via Latent Trajectory Priors: DiffuPrompt tackles the memory drift and topological fracture issues when adapting SAM2 to axial slice propagation by introducing an LLM-guided Neural ODE to predict continuous anatomical latent trajectories under a frozen backbone, which are adaptively fused with observational working memory via a dual-stream spatial gate for high-fidelity single-prompt 3D medical volume segmentation.
Diffusion Model as a Generalized Segmentation Learner: DiGSeg repurposes a pretrained diffusion model into a general-purpose segmentation learner by encoding images and masks into the latent space, injecting multi-scale CLIP text features, and pairing annealed multi-resolution noise scheduling with fast DDIM-trailing denoising, achieving post-processing-free state-of-the-art segmentation and robust cross-domain generalization.
DiffVP: Differential Visual Semantic Prompting for LLM-Based CT Report Generation: Inspired by radiological cognitive subtraction, DiffVP introduces normal chest CT scans as a visual reference prior to model complementary global and local semantic discrepancies in an aligned latent space, which are projected into learnable visual prefix prompts for an LLM to steer report generation without requiring explicit lesion segmentation.
Enhancing Interpretability in CLIP with Optimal Transport-based Submodular Optimization for Ophthalmic Imaging: To address the issues of fragmented activations and noisy backgrounds in ophthalmic CLIP explanations, OTSMDL couples multi-dimensional submodular scoring for candidate initialization with unbalanced optimal transport for global consensus re-ranking, accurately grounding diagnostic lesions and anatomical structures.
FedMental: Topology-Aware Federated Prototype Learning for Polymorphic Multimodal Psychiatry: Targeting the severe non-IID challenges caused by polymorphic disease manifestations in multi-center multimodal psychiatric diagnosis, FedMental introduces Topology-Aware Prototype Refinement (TAPR) via parameter-free clustering on the server alongside a client-side dual-objective alignment (IVMA and CVSC), achieving 46.32% overall accuracy on the real-world Psycare benchmark (+8.83% absolute improvement over FedAvg).
Flexible Control of 3D CT Generation via Text and Semantically-Defined Segmentation Prompts: MedSynV2 introduces a flexible multimodal framework that tokenizes volumetric CT and arbitrary single-lesion/anatomy binary masks into a shared video latent space, combining compound text prompts with a gated-attention DiT to achieve high-resolution 3D CT generation without requiring full-organ annotations.
Follow-Your-Mind: Towards Inversion-Free Brain-Driven Visual Context Synthesis and Editing: MinD-Flow is a pioneering inversion-free framework for brain-driven visual context synthesis and editing based on Rectified Flow, achieving structure-preserving and semantically aligned neural editing in just 5 Euler steps via decoupled flow guidance and neuro-adaptive soft masking.
Foundation-Guided Representation Alignment for Multimodal Medical Image Registration: This paper introduces a foundation-guided representation alignment framework for multimodal medical image registration that transfers cross-modal priors from a 2D vision foundation model (DINOv3) to a lightweight 3D registration network via data reassembly and hierarchical feature fusion, eliminating foundation model dependencies during inference while substantially boosting MR-CT registration accuracy.
From Minimal Clinical Prompts to 3D: Spacing-Aware Prompt Propagation for Multimodal Prostate Lesion Segmentation in bpMRI: Addressing the practical bottlenecks in clinical prostate biparametric MRI (bpMRI)—where routine reports provide only sparse 2D measurements on a single slice and scans exhibit large, anisotropic axial slice thickness—this paper proposes SAPP, which couples adaptive multimodal cross-attention fusion with spacing-decayed, confidence-gated streaming-memory propagation and an explicit stop classifier, extending SAM 2 to reliably lift minimal 2D clinical prompts into coherent 3D lesion masks.
FSD-Net: Foundation-Guided Spatiotemporal Distillation for Video Polyp Segmentation: To alleviate the prohibitive cost of dense pixel-level annotations and eliminate mask hallucinations caused by mucosal occlusions, FSD-Net introduces a foundation-guided spatiotemporal distillation framework with occlusion-aware cycle consistency and momentum prototype re-identification, allowing a lightweight student model to surpass fully supervised methods at 46.7 FPS under sparse annotations.
FST-SAM3: Taming SAM~3 with Frequency-Spatio-Temporal Refinement for Video Polyp Segmentation: To tackle mucosal camouflage, ambiguous boundaries, and error accumulation in video polyp segmentation, FST-SAM3 adapts frozen SAM 3 with encoder-side frequency enhancement and decoder-side spatio-temporal token refinement, adding only 3.7M parameters while establishing new state-of-the-art results under both automatic and prompted protocols.
Generalized Biomedicine Discovery: Addressing the suppression of subtle lesions and rare diseases by dominant normal anatomy in clinical open-world environments, this paper introduces the Generalized Biomedicine Discovery (GBD) benchmark and SCAN, a cognition-inspired plug-and-play vision layer that filters predictable patterns via predictive suppression, excites anomalous salience via surprise energy, and performs complementary accommodation to discover novel concepts while preserving established clinical knowledge.
GeoCFM: Positive-Only Conditional Flow Matching for Mineral Occurrence Sampling: Addressing the absence of reliable negatives and the epistemic non-identifiability of unobserved subsurface geology in mineral exploration, GeoCFM recasts mineral targeting from deterministic score-map prediction into conditional flow matching over continuous spatial point densities, generating multi-hypothesis occurrence distributions without pseudo-negatives.
Histocomponent-driven Universal Model for Virtual Immunohistochemistry Multiplex Staining via Joint Manifold Evolution: To tackle manifold drift caused by external text prompts and the absence of tissue heterogeneity modeling in universal virtual IHC multiplex staining, HUSE introduces Joint Manifold Anchoring, Histocomponent-driven Mixture of Experts, and Representation Conflict Gating within a pixel-level flow matching framework, alongside a pioneering mIF-driven data synthesis paradigm for high-fidelity "one-to-many" multiplex generation.
iMED: A Multi-Endoscope Dataset for Surgical 3D Perception: Addressing the fundamental limitation of surgical vision benchmarks confined to narrow single-trajectory forward motion, iMED introduces the first synchronized dual-stereo-endoscope dataset (340 sequences, ~170K timepoints) and a cross-trajectory evaluation protocol across 23 SOTA algorithms, revealing photometric collapse under held-out viewpoints and the remarkable transferability of multi-domain foundation models.
Implicit Neural Representation for Spherical Harmonics Reconstruction of Motion-Corrupted Fetal Diffusion MRI: Addressing severe motion corruption and single-orientation stripe artifacts in fetal diffusion MRI, this paper proposes the first self-supervised implicit neural representation framework parameterized by continuous spherical harmonics, achieving high-fidelity 3D reconstruction and sharp tractography via physics-based acquisition modeling and through-plane continuity regularization.
MArFE: Multi-Contrast MRI Arbitrary Scale Super-Resolution with Fourier Enhancement: To alleviate the inherent spectral bias of implicit neural representations and insufficient multi-contrast spatial feature fusion, this paper introduces MArFE, a Fourier-enhanced arbitrary-scale multi-contrast MRI super-resolution framework incorporating fixed Fourier basis reparameterization, a non-local Galerkin linear attention operator, and dual-domain losses for continuous-scale high-frequency detail recovery.
MCPNet:Masked Coordinate Pooling-based Attention Network for Medical Landmark Detection: Addressing the loss of spatial position information caused by global pooling in conventional attention mechanisms, MCPNet proposes a Masked Coordinate Pooling (MCP) module that projects CNN features into the coordinate domain using static Cartesian and adaptive polar-like coordinate maps, models anatomical contextual relationships via self-attention, and smoothly reconstructs features back to the convolutional space through mask-wise recalibration, achieving SOTA across multiple 2D/3D medical landmark detection benchmarks.
Mechanistic interventions for explainable digital pathology uncovers adversarial vulnerabilities: Addressing the black-box opacity and vulnerability of digital pathology models to adversarial perturbations, this work presents a mechanistic framework combining sparse autoencoders and causal activation patching to uncover spurious artifact dependencies (e.g., staining variations and scanner noise), and introduces Mechanism-Informed Adversarial Training (MIAT) to surgically regularize vulnerable neural circuits while maintaining clean diagnostic accuracy.
MedCAGD: Context-Aware Gated Decoder for Efficient Medical Image Segmentation: MedCAGD proposes a decoder-centric medical image segmentation framework that systematically regulates skip connection fusion and multi-scale context aggregation through a context-aware gated decoder. At a lightweight computational cost of 30.60M parameters and 5.0 GFLOPs, it consistently outperforms CNN, Transformer, Mamba, and SOTA decoder methods across 11 medical image segmentation benchmarks.
MedRepBench: Benchmarking Structured Understanding of Medical Report Images: MedRepBench introduces an end-to-end benchmark of 1,925 de-identified real-world Chinese medical report images, establishes dual objective (field-level recall) and subjective (LLM judge) evaluation protocols, and demonstrates that a lightweight GRPO reinforcement learning baseline using average field recall as a reward boosts end-to-end VLM recall by 6.14%.
Mimicking Radiologists: A Coarse-to-Fine Framework with Structural Sparse Tokens for Dual-LLM Computed Tomography Report Generation: This paper proposes a coarse-to-fine dual-LLM framework mimicking the visual search workflow of radiologists for 3D computed tomography report generation, combining structural global token screening, abnormality-prompted multi-source local token filtering, and GRPO collaborative alignment to compress visual tokens by 96.1% while boosting clinical efficacy (CE-F1) by 32.0%.
Neuromorphic X-ray Computed Tomography: Presents the first application of neuromorphic sensing to X-ray CT and proposes the Event-enhanced Neural Attenuation Field (ENAF), which couples sparse intensity projections with microsecond asynchronous event streams to resolve angular ambiguities and significantly boost sparse-view reconstruction accuracy and efficiency.
PhenoLIP: Phenotype Guided Medical Vision–Language Pretraining: The authors construct PhenoKG, the first large-scale multimodal knowledge graph linking over 524K image-text pairs to more than 3,000 clinical phenotypes, and propose PhenoLIP, a two-stage pretraining framework that embeds hierarchical ontology structures and distills them into a VLM to dramatically boost zero-shot medical recognition and retrieval.
Proto-Gaussian: MRI Modality Translation Based on Learnable Structural Prototypes and 2D Gaussian Splatting: Proto-Gaussian introduces the first cross-modal MRI translation framework based on explicit 2D Gaussian parametric representations, disentangling anatomical geometry from modality-specific textures via a shared geometric prototype bank and a label-adaptive token refinement module, followed by high-fidelity synthesis via a fully differentiable 2D Gaussian splatting renderer.
PyraE2E: Enhancing End-to-End WSI Analysis via Cross-Scale Super-Resolution: To resolve the critical dilemma in end-to-end whole slide image (WSI) analysis where computational constraints necessitate low-magnification inputs and random sampling—causing severe morphological loss and sparse supervision—PyraE2E leverages the native multi-resolution WSI pyramid to formulate an auxiliary cross-scale super-resolution task, paired with an embedded cluster-score sampling module and a lightweight partitioned encoder, achieving state-of-the-art diagnostic performance under strictly controlled computational cost.
Rethinking Real-World MRI Denoising: Learning from Physical Noise: Addressing the core obstacle in magnetic resonance imaging (MRI) where noise-free reference targets are physically unobtainable, this paper introduces YADO, which directly leverages repeated scans to establish physical Noise2Noise (pN2N), generalizes to single acquisitions via diffusion-based re-noising (ReN2N), and incorporates co-acquired contrast guidance to outperform 17 baselines across 14 benchmarks while matching brute-force physically averaged scans.
Safe Generalization: Mitigating Catastrophic Forgetting in Single-Source Multi-Organ Segmentation via Collaborative Causal Learning: To tackle catastrophic forgetting in single-source domain generalization (CF-SDG) where cross-domain training destroys source domain accuracy, this paper proposes CCL-Net, which severs spurious appearance shortcuts via dual-path causal intervention (front-door orthogonal disentanglement and back-door anatomy-guided local intervention) and preserves source knowledge via dynamic local-global collaborative constraints.
Scaling Whole-Slide Pathology Foundation Model Pretraining with Billions Off-the-Shelf Tokens: To break the dilemma between information collapse in tile-level [CLS] tokens and prohibitive quadratic compute of all spatial tokens in gigapixel pathology, this paper introduces a training-free greedy diversity retention module (TokRet) and a block-sparse long-context pretraining framework (PathTokScale), scaling pretraining to 2 billion tokens with only 10% of the WSI volume required by prior SOTA foundation models.
SCDL: Synergistic Confidence-Dispersion Learning for Semi-Supervised Video Polyp Segmentation: To tackle low polyp-mucosa contrast and error propagation from overconfident pseudo-labels in colonoscopy videos, SCDL introduces prototype-guided adaptive refinement, spectral convex optimization over confidence-dispersion space, and contextual consistency perturbation, creating a closed-loop semi-supervised framework that establishes new SOTA results across all label ratios on SUN-SEG.
SDUM: A Scalable Deep Unrolled Model for Universal Cardiac MRI Reconstruction: Addressing protocol-specific brittleness in clinical cardiac MRI across diverse contrasts, trajectories, and accelerations, SDUM introduces a scalable deep unrolled framework incorporating Restormer image priors, per-cascade coil sensitivity estimation, sampling-aware weighted data consistency, and universal acquisition conditioning, achieving state-of-the-art across all CMRxRecon tracks without fine-tuning while establishing foundational scaling laws.
SeekFlow: Synergizing Radiology and Pathology Foundation Models for Precision Oncology via Knowledge-Guided Evidence Flow: To tackle scale mismatch and topological degeneracy between macro-scale radiology and micro-scale pathology foundation models, SeekFlow builds a clinical prototype-anchored evidence manifold, employs knowledge-guided continuous flow matching to achieve topology-preserving feature transport, and integrates hierarchical evidential aggregation via Dempster-Shafer theory, achieving state-of-the-art performance across tumor grading, staging, and survival prediction.
Short-to-Long Functional Connectivity Transfer via Structure-Aware Latent Diffusion: Addressing the noise in short-scan fMRI and the attenuation of individual phenotypic variance in standard diffusion models, this paper introduces SALD, a structure-aware latent diffusion framework that combines structural connectivity priors with a reward-aligned LoRA adaptation steered by phenotype distillation and one-step denoising anchoring, significantly restoring downstream cognitive prediction while preserving network topology.
SPDA: Efficient Online Test-Time Adaptation for Promptable Medical Segmentation: Addressing performance degradation of general and medical foundation segmentation models under clinical sub-domain shifts, SPDA presents an efficient online test-time adaptation framework that constructs reliable self-supervised learning signals via domain-adaptive low-frequency radial-spectrum perturbation and prompt under-sampling, while updating only a lightweight post-encoder adapter to bypass heavy encoder backpropagation.
SPHERE: From MRI Sampling Mechanisms to Spatial Priors for Generalizable Brain Tumor Segmentation: To overcome the severe domain degradation in brain tumor MRI segmentation caused by purely image-domain models ignoring acquisition-induced structural distortions, SPHERE recovers k-space physical sampling parameters directly from reconstructed scans, constructs a continuous deformation vector field (DVF), and injects it into a lightweight Swin-RetNet via a Deformation Prior Module (DPM) to deliver superior boundary stability, cross-dataset generalization, and computational efficiency.
Steering Diffusion Models via Class-Contrastive Influence for Few-Shot Classification: Addressing the core issue that synthetic data augmentation in few-shot classification prioritizes image realism over downstream utility, this paper proposes Class-Contrastive Influence (C2I) along with theoretical grounding, demonstrating that high-C2I samples are boundary-proximal hard examples, and fine-tunes diffusion models via reinforcement learning using C2I rewards to generate task-informative samples that substantially boost downstream accuracy and noise robustness.
Structured SIR: Efficient and Expressive Importance-Weighted Inference for High-Dimensional Image Registration: To overcome the limitations of variational inference in capturing multi-modal posteriors and high-dimensional spatial correlations in dense 3D image registration, this paper introduces Structured SIR—a framework combining Sampled Importance Resampling with a compact "low-rank plus locally sparse Cholesky precision factor" covariance parameterization that delivers expressive, multi-modal, and well-calibrated probabilistic inference across 1.6 million displacement dimensions.
Two-Way Street: Efficient VSLAM using Collaborative In-Sensor and Off-Sensor processing: This paper introduces a collaborative in-sensor and off-sensor VSLAM architecture based on the SCAMP-7 Pixel Processor Array (PPA), delegating high-speed point-feature detection and tracking to in-pixel SIMD compute at up to 2000 Hz while using back-end camera pose feedback to guide sensor-level rotation-invariant matching, cutting per-frame data transfer by over 1000× and per-frame energy consumption by approximately 20×.
Unified Multi-plane Autoregressive Diffusion for 3D Multi-Contrast MRI Synthesis: The paper introduces the unified Multi-Plane Autoregressive Diffusion (MPAD) framework, which couples an isotropic 3D latent space with 2D masked-slice diffusion training, along with cross-orthogonal-plane autoregressive synthesis and prior propagation, synthesizing volumetrically coherent multi-contrast brain MRI scans with 2D computational efficiency.
UniH3: Unifying Hierarchical Homogeneity and Heterogeneity for All-in-One Medical Image Restoration: UniH3 establishes the first all-in-one medical image restoration framework that systematically unifies hierarchical homogeneity and heterogeneity, distilling and retrieving cross-task anatomical priors via Hierarchical Homogeneity Memory (H2M) and Homogeneity-Guided Attention (HGA) while resolving task- and sample-level gradient conflicts via Hierarchical Heterogeneity Balancer (H2B), achieving new SOTA performance on MedIR-2D-500K and MedIR-3D-3K benchmarks.
Variational Patch Gating for Training-Free Few-Shot Classification: Addressing the collapse of global CLS features and the prohibitive computational overhead of optimal transport patch solvers under domain shift, this paper proposes Variational Patch Gating (VPG), formulating patch classification as free-energy minimization to analytically derive closed-form sigmoid gating and bounded softplus evidence accumulation, dramatically outperforming prior training-free methods across cross-domain benchmarks.
Verifying Cancer Segmentation in Vision Transformers via Internal Concepts: By demonstrating that Vision Transformers exhibit a latent "Feeling of Error" (characterized by fewer active internal concepts with weaker magnitudes during prediction failures), this paper extracts hierarchical Sparse Autoencoder concepts across network depths to achieve highly accurate, interpretable failure detection and patch-level false positive correction without sacrificing segmentation quality.
Wavelet-Driven Cross-Domain Consistency for Mixed-Supervised 3D Tumor Segmentation: To tackle CT reconstruction-kernel heterogeneity and multi-source domain shifts, this paper proposes WCC4MS, a wavelet-driven cross-domain consistency framework for mixed-supervised 3D tumor segmentation that synthesizes smooth/sharp variants via 3D wavelet modulation and regularizes a dual-branch network with cross-task KL consistency and domain-aware objectives.
XSurfer: Reconstructing surface meshes of cerebral and cerebellar cortex from diverse MRI data using untrained neural networks: XSurfer repairs sulcal connections in tissue segmentations and extracts initial surfaces, then optimizes a randomly initialized network separately for each subject to deform the white surface into a pial surface, achieving pial ASSD of 0.303/0.241 mm on UKB and dHCP infants; the surface-deformation network requires no pretraining, but this does not apply to the entire pipeline, which includes segmentation.
Y-diff: Structure-Texture Decoupled Diffusion Distillation for H&E-to-pCLE Translation: Y-diff proposes a structure-texture decoupled teacher-student diffusion distillation framework that leverages DiffAE semantic flow matching to transfer pCLE optical textures and grayscale SPNet spatial priors to anchor local cellular topology, substantially eliminating unpaired translation artifacts and reducing downstream cell counting MAPE from 55.77% to 13.84%.