arXiv subjects

Yalin Zheng

Publications and source records attributed to Yalin Zheng.

At least 19 recordsLinked to original sources

arXiv2026

PARAGraph: Pathology-Anatomy-Aware Hierarchical Graph for Diabetic Retinopathy Grading

Diabetic retinopathy (DR) remains a leading cause of vision loss among working-age adults worldwide, making reliable severity grading clinically important. Despite strong performance, most deep models formulate DR grading as image-level classification and do not explicitly model clinically grounded evidence, such as lesion types and spatial relations. In this paper, we propose PARAGraph, a Pathology-Anatomy-Aware Hierarchical Graph framework for DR grading. PARAGraph represents each image as a three-level hierarchical graph with lesion-level nodes, intermediate category and region nodes, and global anatomical and semantic nodes. To incorporate medical priors into nodes, we construct an optic disc-fovea-anchored coordinate frame that provides a scale- and rotation-normalized retinal reference system. Within this frame, lesion nodes are encoded with category, normalized area, and anatomical coordinates. To mitigate noisy lesion segmentation, PARAGraph uses a dual-fusion strategy that introduces global visual context into a graph semantic node and a decision-level prediction branch, improving robustness when lesion evidence is unreliable. Extensive experiments on Messidor-2, APTOS, and DDR show that PARAGraph achieves consistent DR grading performance over state-of-the-art methods. Interpretability and robustness analyses further demonstrate that its predictions are clinically grounded, closely associated with lesion evidence and robust to lesion segmentation noise.

cs.CV↗

arXiv2026

P2E-VQ: ECG-linked representation augmentation for PPG via discrete patch retrieval

Photoplethysmography (PPG) is widely used in consumer wearables because of its low cost and ease of acquisition. However, unlike electrocardiography (ECG), PPG measures peripheral pulse dynamics rather than cardiac electrical activity, limiting its ability to predict cardiac conditions that rely on ECG-specific morphological cues. Existing methods attempt to bridge this gap by reconstructing ECG signals from PPG signals. However, this inverse mapping is inherently ill-posed, and faithful waveform reconstruction does not necessarily translate into improved downstream performance. To address this challenge, we propose P2E-VQ, a retrieval-augmented framework that replaces ECG waveform reconstruction with ECG-linked representation retrieval. Specifically, P2E-VQ converts PPG patches into discrete tokens and retrieves ECG-linked information from a memory bank constructed exclusively from the training data. This process augments PPG representations while requiring only PPG signals during inference. Extensive experiments on five public datasets covering six downstream tasks, including clinical endpoint prediction and affective state recognition, demonstrate that P2E-VQ consistently outperforms pretrained baselines under a unified frozen-feature linear-probing protocol.

cs.LG↗

arXiv2026

Beyond Visual Grasping: Benchmarking Complex Grasping from Detection to Execution

Robust robotic grasping remains a fundamental challenge for complex real-world applications. Recent advances in large-scale models demonstrate promising capabilities for reasoning in robotic tasks. However, existing benchmarks for grasping primarily focus on isolated, visual-based grasp pose detection, failing to capture the complexity of grasping tasks that require multi-step reasoning and semantic understanding during execution. To address this gap, we propose GCA-Bench, a benchmark featuring challenging \textit{grasping with complex action} scenarios that involve both scene-level reasoning and semantic constraints. GCA-Bench enables the evaluation of recent large foundation models under the same settings. To demonstrate the effectiveness of our new benchmark, we implement a diverse set of baselines, ranging from traditional grasp detection pipelines to end-to-end learning methods. Empirical studies achieve success rates below 70\% on complex grasping scenarios, underscoring critical limitations. In addition, we propose new evaluation metrics, analyze critical failure models, and provide insights to guide the development of more robust and generalizable grasping strategies.

cs.RO↗

arXiv2026

HadBalance: A Plug-and-Play Unified Global Geometric Prior Framework for Generalizable Biomedical Segmentation

Precise biomedical image segmentation is crucial for clinical diagnosis. Geometric cues (e.g., boundary, shape, and topology) can improve structural consistency, yet most are task-specific and lack a unified geometric foundation that generalizes across organs and modalities. We are motivated by the observation that several medical segmentation targets can be approximated as globally near-convex shapes. A convex region is one in which any two interior points can be connected by a line segment entirely contained within the region. In practice, medical targets may exhibit small local concavities or boundary irregularities; we refer to such globally convex-like shapes as near-convex. Motivated by this, we derive Hadwiger Shape Priors from Hadwiger's theorem as an interpretable global regularizer using three 2D measures: area A, perimeter P, and Euler characteristic chi, enabling transfer across organs and modalities. However, because medical datasets are shape-heterogeneous, enforcing near-convex priors uniformly can over-regularize non-convex anatomy with significant concavities, washing out concavities and fine details and degrading segmentation accuracy. To address this challenge, we propose Conflict-Aware Objective Balancing (CAOB), which integrates shape priors with segmentation in a gradient-aware manner. For each prior, CAOB removes only the gradient component that conflicts with segmentation while preserving the remaining aligned component, and adaptively regulates objective influences to prevent prior dominance. This enables stable use of shape priors on shape-heterogeneous data without erasing genuine concavities or fine structural details. We call this plug-and-play framework HadBalance.

cs.CV↗

arXiv2026

Disentangled Fine-Grained Prototype Learning for Incomplete Image-Tabular Classification

The missing-modality problem poses a significant challenge in image-tabular multimodal learning across a wide range of multimedia applications, including product understanding, recommendation systems, and medical diagnosis. This challenge is particularly pronounced when the two modalities are highly heterogeneous, as images and tabular attributes differ substantially in their semantic granularity and data distributions. Existing methods learn modality-invariant representations through disentanglement and alignment over global token-averaged features, capturing only coarse cross-modal consistency and overlooking fine-grained semantic and distributional misalignment, which hampers the exploitation of complementary cues under missing modalities. To address this, we propose DFPL, a novel framework for fine-grained prototype learning. Specifically, Shared-Specific Prototype Modeling (SSPM) extracts compact and diverse shared and modality-specific prototypes, and further performs prototype-level disentanglement to suppress redundant intra-modality correlations. Additionally, we propose a Prototype-guided Fine-grained Alignment (PFA) module that jointly enforces prototype-level distribution matching and prototype-to-class semantic alignment within a unified prototype space, thereby preserving both fine-grained distributional and semantic consistency across modalities. We further introduce a Class-aware Multi-scale Aggregation (CMA) module to adaptively aggregate shared semantics and modality-specific characteristics from global and prototype levels for robust predictions. Extensive experiments on three diverse image-tabular benchmarks demonstrate the superiority of our method compared to the previous approaches under various missing-modality settings. Code will be made publicly available.

cs.CV↗

arXiv2026

Empowering Locally Deployable Medical Agent via State Enhanced Logical Skills for FHIR-based Clinical Tasks

While Large Language Models demonstrate immense potential as proactive Medical Agents, their real-world deployment is severely bottlenecked by data scarcity under privacy constraints. To overcome this, we propose State-Enhanced Logical-Skill Memory (SELSM), a training-free framework that distills simulated clinical trajectories into entity-agnostic operational rules within an abstract skill space. During inference, a Query-Anchored Two-Stage Retrieval mechanism dynamically fetches these entity-agnostic logical priors to guide the agent's step-by-step reasoning, effectively resolving the state polysemy problem. Evaluated on MedAgentBench -- the only authoritative high-fidelity virtual EHR sandbox benchmarked with real clinical data -- SELSM substantially elevates the zero-shot capabilities of locally deployable foundation models (30B--32B parameters). Notably, on the Qwen3-30B-A3B backbone, our framework completely eliminates task chain breakdowns to achieve a 100\% completion rate, boosting the overall success rate by an absolute 22.67\% and significantly outperforming existing memory-augmented baselines. This study demonstrates that equipping models with a dynamically updatable, state-enhanced cognitive scaffold is a privacy-preserving and computationally efficient pathway for local adaptation of AI agents to clinical information systems. While currently validated on FHIR-based EHR interactions as an initial step, the entity-agnostic design of SELSM provides a principled foundation toward broader clinical deployment.

cs.AI↗

arXiv2026

A WDLoRA-Based Multimodal Generative Framework for Clinically Guided Corneal Confocal Microscopy Image Synthesis in Diabetic Neuropathy

Corneal Confocal Microscopy (CCM) is a sensitive tool for assessing small-fiber damage in Diabetic Peripheral Neuropathy (DPN), yet the development of robust, automated deep learning-based diagnostic models is limited by scarce labelled data and fine-grained variability in corneal nerve morphology. Although Artificial Intelligence (AI)-driven foundation generative models excel at natural image synthesis, they often struggle in medical imaging due to limited domain-specific training, compromising the anatomical fidelity required for clinical analysis. To overcome these limitations, we propose a Weight-Decomposed Low-Rank Adaptation (WDLoRA)-based multimodal generative framework for clinically guided CCM image synthesis. WDLoRA is a parameter-efficient fine-tuning (PEFT) mechanism that decouples magnitude and directional weight updates, enabling foundation generative models to independently learn the orientation (nerve topology) and intensity (stromal contrast) required for medical realism. By jointly conditioning on nerve segmentation masks and disease-specific clinical prompts, the model synthesises anatomically coherent images across the DPN spectrum (Control, T1NoDPN, T1DPN). A comprehensive three-pillar evaluation demonstrates that the proposed framework achieves state-of-the-art visual fidelity (Fr\'echet Inception Distance (FID): 5.18) and structural integrity (Structural Similarity Index Measure (SSIM): 0.630), significantly outperforming GAN and standard diffusion baselines. Crucially, the synthetic images preserve gold-standard clinical biomarkers and are statistically equivalent to real patient data. When used to train automated diagnostic models, the synthetic dataset improves downstream diagnostic accuracy by 2.1% and segmentation performance by 2.2%, validating the framework's potential to alleviate data bottlenecks in medical AI.

cs.CV↗

arXiv2026

Leveraging Persistence Image to Enhance Robustness and Performance in Curvilinear Structure Segmentation

Segmenting curvilinear structures in medical images is essential for analyzing morphological patterns in clinical applications. Integrating topological properties, such as connectivity, improves segmentation accuracy and consistency. However, extracting and embedding such properties - especially from Persistence Diagrams (PD) - is challenging due to their non-differentiability and computational cost. Existing approaches mostly encode topology through handcrafted loss functions, which generalize poorly across tasks. In this paper, we propose PIs-Regressor, a simple yet effective module that learns persistence image (PI) - finite, differentiable representations of topological features - directly from data. Together with Topology SegNet, which fuses these features in both downsampling and upsampling stages, our framework integrates topology into the network architecture itself rather than auxiliary losses. Unlike existing methods that depend heavily on handcrafted loss functions, our approach directly incorporates topological information into the network structure, leading to more robust segmentation. Our design is flexible and can be seamlessly combined with other topology-based methods to further enhance segmentation performance. Experimental results show that integrating topological features enhances model robustness, effectively handling challenges like overexposure and blurring in medical imaging. Our approach on three curvilinear benchmarks demonstrate state-of-the-art performance in both pixel-level accuracy and topological fidelity.

cs.CV↗

arXiv2026

StealthMark: Harmless and Stealthy Ownership Verification for Medical Segmentation via Uncertainty-Guided Backdoors

Annotating medical data for training AI models is often costly and limited due to the shortage of specialists with relevant clinical expertise. This challenge is further compounded by privacy and ethical concerns associated with sensitive patient information. As a result, well-trained medical segmentation models on private datasets constitute valuable intellectual property requiring robust protection mechanisms. Existing model protection techniques primarily focus on classification and generative tasks, while segmentation models-crucial to medical image analysis-remain largely underexplored. In this paper, we propose a novel, stealthy, and harmless method, StealthMark, for verifying the ownership of medical segmentation models under black-box conditions. Our approach subtly modulates model uncertainty without altering the final segmentation outputs, thereby preserving the model's performance. To enable ownership verification, we incorporate model-agnostic explanation methods, e.g. LIME, to extract feature attributions from the model outputs. Under specific triggering conditions, these explanations reveal a distinct and verifiable watermark. We further design the watermark as a QR code to facilitate robust and recognizable ownership claims. We conducted extensive experiments across four medical imaging datasets and five mainstream segmentation models. The results demonstrate the effectiveness, stealthiness, and harmlessness of our method on the original model's segmentation performance. For example, when applied to the SAM model, StealthMark consistently achieved ASR above 95% across various datasets while maintaining less than a 1% drop in Dice and AUC scores, significantly outperforming backdoor-based watermarking methods and highlighting its strong potential for practical deployment. Our implementation code is made available at: https://github.com/Qinkaiyu/StealthMark.

cs.CV↗

arXiv2025

SCR2-ST: Combine Single Cell with Spatial Transcriptomics for Efficient Active Sampling via Reinforcement Learning

Spatial transcriptomics (ST) is an emerging technology that enables researchers to investigate the molecular relationships underlying tissue morphology. However, acquiring ST data remains prohibitively expensive, and traditional fixed-grid sampling strategies lead to redundant measurements of morphologically similar or biologically uninformative regions, thus resulting in scarce data that constrain current methods. The well-established single-cell sequencing field, however, could provide rich biological data as an effective auxiliary source to mitigate this limitation. To bridge these gaps, we introduce SCR2-ST, a unified framework that leverages single-cell prior knowledge to guide efficient data acquisition and accurate expression prediction. SCR2-ST integrates a single-cell guided reinforcement learning-based (SCRL) active sampling and a hybrid regression-retrieval prediction network SCR2Net. SCRL combines single-cell foundation model embeddings with spatial density information to construct biologically grounded reward signals, enabling selective acquisition of informative tissue regions under constrained sequencing budgets. SCR2Net then leverages the actively sampled data through a hybrid architecture combining regression-based modeling with retrieval-augmented inference, where a majority cell-type filtering mechanism suppresses noisy matches and retrieved expression profiles serve as soft labels for auxiliary supervision. We evaluated SCR2-ST on three public ST datasets, demonstrating SOTA performance in both sampling efficiency and prediction accuracy, particularly under low-budget scenarios. Code is publicly available at: https://github.com/hrlblab/SCR2ST

cs.CV↗

arXiv2025

EyeAgent: An Agentic AI System for Multimodal Clinical Decision Support in Ophthalmology

Artificial intelligence has shown promise in medical imaging, yet most existing systems lack flexibility, interpretability, and adaptability - challenges especially pronounced in ophthalmology, where diverse imaging modalities are essential. We present EyeAgent, the first agentic AI framework for comprehensive and interpretable clinical decision support in ophthalmology. Using a large language model (DeepSeek-V3) as its central reasoning engine, EyeAgent interprets user queries and dynamically orchestrates 53 validated ophthalmic tools across 23 imaging modalities for diverse tasks including classification, segmentation, detection, image/report generation, and quantitative analysis. Stepwise ablation analysis demonstrated a progressive improvement in diagnostic accuracy, rising from a baseline of 69.71% (using only 5 general tools) to 80.79% when the full suite of 53 specialized tools was integrated. In an expert rating study on 200 real-world clinical cases, EyeAgent achieved 93.7% tool selection accuracy and received expert ratings of more than 88% across accuracy, completeness, safety, reasoning, and interpretability. In human-AI collaboration, EyeAgent matched or exceeded the performance of senior ophthalmologists and, when used as an assistant, improved overall diagnostic accuracy by 18.51% and report quality scores by 19%, with the greatest benefit observed among junior ophthalmologists. These findings establish EyeAgent as a scalable and trustworthy AI framework for ophthalmology and provide a blueprint for modular, multimodal, and clinically aligned next-generation AI systems.

cs.HC↗

arXiv2025

PSScreen V2: Partially Supervised Multiple Retinal Disease Screening

In this work, we propose PSScreen V2, a partially supervised self-training framework for multiple retinal disease screening. Unlike previous methods that rely on fully labelled or single-domain datasets, PSScreen V2 is designed to learn from multiple partially labelled datasets with different distributions, addressing both label absence and domain shift challenges. To this end, PSScreen V2 adopts a three-branch architecture with one teacher and two student networks. The teacher branch generates pseudo labels from weakly augmented images to address missing labels, while the two student branches introduce novel feature augmentation strategies: Low-Frequency Dropout (LF-Dropout), which enhances domain robustness by randomly discarding domain-related low-frequency components, and Low-Frequency Uncertainty (LF-Uncert), which estimates uncertain domain variability via adversarially learned Gaussian perturbations of low-frequency statistics. Extensive experiments on multiple in-domain and out-of-domain fundus datasets demonstrate that PSScreen V2 achieves state-of-the-art performance and superior domain generalization ability. Furthermore, compatibility tests with diverse backbones, including the vision foundation model DINOv2, as well as evaluations on chest X-ray datasets, highlight the universality and adaptability of the proposed framework. The codes are available at https://github.com/boyiZheng99/PSScreen_V2.

cs.CV↗

arXiv2025

RetiBridge: Bridging Quantitative Retinal Biomarkers and Qualitative Diagnosis with a Knowledge-Guided Multimodal Large Language Model

Retinal biomarkers captured by color fundus photography and optical coherence tomography provide clinically valuable evidence for both ocular and systemic diseases. Multimodal large language models (MLLMs) have shown promise for retinal image interpretation, yet existing ophthalmic models rarely quantify these clinically relevant biomarkers or explicitly translate their measurements into qualitative, evidence-grounded diagnostic conclusions. To address this gap, we introduce RetiBridge, a knowledge-guided multimodal large language model that jointly analyzes color fundus photography (CFP), optical coherence tomography (OCT), and text, explicitly bridging quantitative retinal biomarkers to qualitative clinical sub-inferences and coherent diagnostic conclusions. RetiBridge combines knowledge-guided instruction generation, OCT-biomarker alignment, and supervised multimodal instruction tuning to learn a biomarker-grounded quantitative-to-qualitative diagnostic pathway. Using 15,611 paired CFP-OCT samples from UK Biobank with 31 OCT and 6 CFP biomarkers, we construct the Grounded Ophthalmic Understanding benchmark to evaluate diagnostic classification, report generation quality, and fine-grained clinical quality. Despite using only LoRA-based fine-tuning of a 7B-parameter Qwen2 backbone, RetiBridge outperforms all evaluated open-source 7B and 32B baselines, achieving the highest quantitative accuracy, evidence grounding, coverage completeness, and BERTScore, while surpassing OpenAI o3 on these key biomarker-grounded metrics. Our code and data are released in the RetiBridge repository.

cs.AI↗

arXiv2025

A Frequency-Aware Self-Supervised Learning for Ultra-Wide-Field Image Enhancement

Ultra-Wide-Field (UWF) retinal imaging has revolutionized retinal diagnostics by providing a comprehensive view of the retina. However, it often suffers from quality-degrading factors such as blurring and uneven illumination, which obscure fine details and mask pathological information. While numerous retinal image enhancement methods have been proposed for other fundus imageries, they often fail to address the unique requirements in UWF, particularly the need to preserve pathological details. In this paper, we propose a novel frequency-aware self-supervised learning method for UWF image enhancement. It incorporates frequency-decoupled image deblurring and Retinex-guided illumination compensation modules. An asymmetric channel integration operation is introduced in the former module, so as to combine global and local views by leveraging high- and low-frequency information, ensuring the preservation of fine and broader structural details. In addition, a color preservation unit is proposed in the latter Retinex-based module, to provide multi-scale spatial and frequency information, enabling accurate illumination estimation and correction. Experimental results demonstrate that the proposed work not only enhances visualization quality but also improves disease diagnosis performance by restoring and correcting fine local details and uneven intensity. To the best of our knowledge, this work is the first attempt for UWF image enhancement, offering a robust and clinically valuable tool for improving retinal disease management.

cs.CV↗

arXiv2025

GLCP: Global-to-Local Connectivity Preservation for Tubular Structure Segmentation

Accurate segmentation of tubular structures, such as vascular networks, plays a critical role in various medical domains. A remaining significant challenge in this task is structural fragmentation, which can adversely impact downstream applications. Existing methods primarily focus on designing various loss functions to constrain global topological structures. However, they often overlook local discontinuity regions, leading to suboptimal segmentation results. To overcome this limitation, we propose a novel Global-to-Local Connectivity Preservation (GLCP) framework that can simultaneously perceive global and local structural characteristics of tubular networks. Specifically, we propose an Interactive Multi-head Segmentation (IMS) module to jointly learn global segmentation, skeleton maps, and local discontinuity maps, respectively. This enables our model to explicitly target local discontinuity regions while maintaining global topological integrity. In addition, we design a lightweight Dual-Attention-based Refinement (DAR) module to further improve segmentation quality by refining the resulting segmentation maps. Extensive experiments on both 2D and 3D datasets demonstrate that our GLCP achieves superior accuracy and continuity in tubular structure segmentation compared to several state-of-the-art approaches. The source codes will be available at https://github.com/FeixiangZhou/GLCP.

cs.CV↗

arXiv2025

Parameterized Diffusion Optimization enabled Autoregressive Ordinal Regression for Diabetic Retinopathy Grading

As a long-term complication of diabetes, diabetic retinopathy (DR) progresses slowly, potentially taking years to threaten vision. An accurate and robust evaluation of its severity is vital to ensure prompt management and care. Ordinal regression leverages the underlying inherent order between categories to achieve superior performance beyond traditional classification. However, there exist challenges leading to lower DR classification performance: 1) The uneven distribution of DR severity levels, characterized by a long-tailed pattern, adds complexity to the grading process. 2)The ambiguity in defining category boundaries introduces additional challenges, making the classification process more complex and prone to inconsistencies. This work proposes a novel autoregressive ordinal regression method called AOR-DR to address the above challenges by leveraging the clinical knowledge of inherent ordinal information in DR grading dataset settings. Specifically, we decompose the DR grading task into a series of ordered steps by fusing the prediction of the previous steps with extracted image features as conditions for the current prediction step. Additionally, we exploit the diffusion process to facilitate conditional probability modeling, enabling the direct use of continuous global image features for autoregression without relearning contextual information from patch-level features. This ensures the effectiveness of the autoregressive process and leverages the capabilities of pre-trained large-scale foundation models. Extensive experiments were conducted on four large-scale publicly available color fundus datasets, demonstrating our model's effectiveness and superior performance over six recent state-of-the-art ordinal regression methods. The implementation code is available at https://github.com/Qinkaiyu/AOR-DR.

cs.CV↗

arXiv2025

Robust Incomplete-Modality Alignment for Ophthalmic Disease Grading and Diagnosis via Labeled Optimal Transport

Multimodal ophthalmic imaging-based diagnosis integrates color fundus image with optical coherence tomography (OCT) to provide a comprehensive view of ocular pathologies. However, the uneven global distribution of healthcare resources often results in real-world clinical scenarios encountering incomplete multimodal data, which significantly compromises diagnostic accuracy. Existing commonly used pipelines, such as modality imputation and distillation methods, face notable limitations: 1)Imputation methods struggle with accurately reconstructing key lesion features, since OCT lesions are localized, while fundus images vary in style. 2)distillation methods rely heavily on fully paired multimodal training data. To address these challenges, we propose a novel multimodal alignment and fusion framework capable of robustly handling missing modalities in the task of ophthalmic diagnostics. By considering the distinctive feature characteristics of OCT and fundus images, we emphasize the alignment of semantic features within the same category and explicitly learn soft matching between modalities, allowing the missing modality to utilize existing modality information, achieving robust cross-modal feature alignment under the missing modality. Specifically, we leverage the Optimal Transport for multi-scale modality feature alignment: class-wise alignment through predicted class prototypes and feature-wise alignment via cross-modal shared feature transport. Furthermore, we propose an asymmetric fusion strategy that effectively exploits the distinct characteristics of OCT and fundus modalities. Extensive evaluations on three large ophthalmic multimodal datasets demonstrate our model's superior performance under various modality-incomplete scenarios, achieving Sota performance in both complete modality and inter-modality incompleteness conditions. Code is available at https://github.com/Qinkaiyu/RIMA

cs.CV↗

arXiv2025

Benchmarking Large Multimodal Models for Ophthalmic Visual Question Answering with OphthalWeChat

Purpose: To develop a bilingual multimodal visual question answering (VQA) benchmark for evaluating VLMs in ophthalmology. Methods: Ophthalmic image posts and associated captions published between January 1, 2016, and December 31, 2024, were collected from WeChat Official Accounts. Based on these captions, bilingual question-answer (QA) pairs in Chinese and English were generated using GPT-4o-mini. QA pairs were categorized into six subsets by question type and language: binary (Binary_CN, Binary_EN), single-choice (Single-choice_CN, Single-choice_EN), and open-ended (Open-ended_CN, Open-ended_EN). The benchmark was used to evaluate the performance of three VLMs: GPT-4o, Gemini 2.0 Flash, and Qwen2.5-VL-72B-Instruct. Results: The final OphthalWeChat dataset included 3,469 images and 30,120 QA pairs across 9 ophthalmic subspecialties, 548 conditions, 29 imaging modalities, and 68 modality combinations. Gemini 2.0 Flash achieved the highest overall accuracy (0.548), outperforming GPT-4o (0.522, P < 0.001) and Qwen2.5-VL-72B-Instruct (0.514, P < 0.001). It also led in both Chinese (0.546) and English subsets (0.550). Subset-specific performance showed Gemini 2.0 Flash excelled in Binary_CN (0.687), Single-choice_CN (0.666), and Single-choice_EN (0.646), while GPT-4o ranked highest in Binary_EN (0.717), Open-ended_CN (BLEU-1: 0.301; BERTScore: 0.382), and Open-ended_EN (BLEU-1: 0.183; BERTScore: 0.240). Conclusions: This study presents the first bilingual VQA benchmark for ophthalmology, distinguished by its real-world context and inclusion of multiple examinations per patient. The dataset reflects authentic clinical decision-making scenarios and enables quantitative evaluation of VLMs, supporting the development of accurate, specialized, and trustworthy AI systems for eye care.

cs.CV↗