Medico 2025: Visual Question Answering for Gastrointestinal Imaging

2025-08-14

AI Summary

The Medico 2025 Challenge addresses explainable visual question answering (VQA) for gastrointestinal (GI) endoscopic images, aiming to enhance AI’s clinical credibility. Existing medical VQA models lack clinically plausible explanations. To address this, we propose a multimodal explainability evaluation framework grounded in expert clinician assessment and introduce Kvasir-VQA-x1—a large-scale, high-fidelity benchmark comprising 6,500 endoscopic images and 158,000 question-answer-explanation triples. Methodologically, we design an end-to-end deep learning architecture integrating a vision encoder, a language understanding module, and an XAI-driven explanation generator that jointly produces answers and clinically coherent, multimodal (text + saliency map) explanations. Our key contributions are: (1) the first GI endoscopy-specific, high-fidelity VQA benchmark; (2) a quantitative, medicine-oriented explainability evaluation paradigm validated by domain experts; and (3) an open-source, clinically deployable multimodal VQA system.

Technology Category

• Computer Vision: Multi-modal Vision
• Humans and AI: Explainable AI (XAI) for Human Understanding
• Machine Learning: Multimodal Learning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Web Mining and Content Analysis: Community question answering
• Security and Privacy: Data transparency and provenance

Abstract

The Medico 2025 challenge addresses Visual Question Answering (VQA) for Gastrointestinal (GI) imaging, organized as part of the MediaEval task series. The challenge focuses on developing Explainable Artificial Intelligence (XAI) models that answer clinically relevant questions based on GI endoscopy images while providing interpretable justifications aligned with medical reasoning. It introduces two subtasks: (1) answering diverse types of visual questions using the Kvasir-VQA-x1 dataset, and (2) generating multimodal explanations to support clinical decision-making. The Kvasir-VQA-x1 dataset, created from 6,500 images and 159,549 complex question-answer (QA) pairs, serves as the benchmark for the challenge. By combining quantitative performance metrics and expert-reviewed explainability assessments, this task aims to advance trustworthy Artificial Intelligence (AI) in medical image analysis. Instructions, data access, and an updated guide for participation are available in the official competition repository: https://github.com/simula/MediaEval-Medico-2025

Problem

• Develop XAI models for GI imaging VQA
• Answer clinical questions with interpretable justifications
• Advance trustworthy AI in medical image analysis

Innovation

• Explainable AI models for GI imaging
• Multimodal explanations for clinical decisions
• Kvasir-VQA-x1 dataset benchmark