IRIS: An Intelligent Vision-Language System for Ocular Surface Diseases via Topic Tree and Scene-Driven VQA Generation

2026-07-05

AI Summary

This work addresses the scarcity of high-quality multimodal instruction data that limits the clinical reasoning capabilities of large vision-language models in the diagnosis and treatment of ocular surface diseases. To overcome this challenge, the authors propose IRIS, a novel system that introduces a Theme Discovery Tree (TFT) structure to align anatomical and pathological concepts. IRIS leverages a scenario-driven strategy to generate role-adaptive clinical dialogues, enabling fine-grained understanding of ocular surface conditions. Through hierarchical theme trees, scenario-driven visual question answering (VQA) generation, and structured multimodal instruction tuning, IRIS substantially outperforms existing general-purpose and specialized medical vision-language models—even those with up to 34 billion parameters—on ocular surface disease VQA tasks. The results demonstrate that structured knowledge injection is more effective than mere parameter scaling and enable efficient deployment on mobile devices.

Abstract

While Large Vision-Language Models (VLMs) demonstrate remarkable generic capabilities, their clinical reasoning in specialized domains like ocular surface diseases (OSDs) is severely hindered by a paucity of high-fidelity, multimodal instruction-tuning data. To dismantle this data bottleneck, we introduce IRIS, an Intelligent Recognition and Interaction System tailored for fine-grained OSD understanding via external eye photography. First, we curate IRIS-120K, the largest and most comprehensive OSD visual question-answering (VQA) dataset to date. Crucially, to overcome the semantic shallowness of conventional image-caption pairs, we propose a synergistic data generation paradigm to explicitly inject clinical priors. Our data engine operates via a dual-branch framework: 1) a Topic Finding Tree (TFT) that hierarchically anchors visual features to precise anatomical and pathological concepts, enforcing rigorous medical deduction logic; and 2) a Scene-driven strategy that synthesizes role-adaptive clinical dialogues to ensure pragmatic generalization. By explicitly aligning a compact 4B-parameter VLM on this structurally enriched corpus, IRIS achieves state-of-the-art performance, comprehensively outperforming both generalist and specialized medical VLMs with up to 34B parameters. Our findings underscore that structured knowledge injection profoundly prevails over sheer parameter scaling, unlocking the potential for resource-efficient, expert-level AI deployment on mobile edge devices for scalable OSD screening. Code, datasets, and model weights will be publicly released by this repo.

Problem

• ocular surface diseases
• vision-language models
• multimodal data
• clinical reasoning
• data bottleneck

Innovation

• Topic Finding Tree
• Scene-driven VQA
• Ocular Surface Diseases
• Structured Knowledge Injection
• Vision-Language Model