MEDLAYXPLAIN: Benchmarking the Expert-Lay Gap in Medical Vision-Language Models

2026-06-19

AI Summary

This study addresses the challenge that existing medical vision-language models (VLMs) struggle to generate layperson-friendly image descriptions, hindering effective physician–patient communication. To bridge this gap, the authors introduce MedLayXPlain-122K, the first large-scale benchmark for medical lay language generation, and propose HOVER—a novel pipeline integrating patient-centered vocabulary mapping, large language model–guided constrained rewriting, and cross-model visual verification—to produce hallucination-free, semantically equivalent expert–lay description pairs. The work also formally defines and quantifies the “expert–lay gap” for the first time, presenting a three-tier semantic alignment framework grounded in the UMLS ontology and a lightweight clinical alignment evaluator, MedLayEval (3B), distilled from a 27B model. Evaluations across 33 VLMs reveal significant degradation in lay language capability among medical VLMs and insufficient clinical accuracy in general-purpose VLMs, highlighting their limitations in patient-facing communication scenarios.

Abstract

Medical Vision-Language Models (Med-VLMs) achieve strong expert-level performance, yet their ability to generate patient-accessible descriptions remains underexplored. With the 21st Century Cures Act now mandating immediate patient access to diagnostic imaging results, evaluating whether Med-VLMs can bridge this Expert-Lay Gap is both urgent and clinically consequential for patient education and shared decision-making. To this end, we introduce MedLayXPlain, the first large-scale multimodal benchmark and evaluation framework for Medical Lay Language Generation (MLLG). MedLayXPlain-122K provides 122,789 region-grounded samples across 8 imaging modalities from 12 publicly available source datasets, each comprising a medical image with paired expert and lay captions anchored in a three-level Unified Medical Language System (UMLS) ontology hierarchy spanning 7 semantic groups, 43 semantic types, and 2,411 medical concepts. Lay captions are constructed via Hierarchical Ontology-Verified Refinement (HOVER), a three-step pipeline combining patient-centric vocabulary mapping, LLM-based constrained rewriting, and cross-model visual verification to enforce semantic equivalence while preventing hallucination. We further introduce MedLayEval, a lightweight 3B evaluator distilled from a 27B verifier that scores expert-lay alignment across five clinically grounded attributes, addressing the poor correlation between standard NLG metrics and clinical judgment. Benchmarking 33 VLMs on MedLayXPlain-122K reveals a systematic Expert-Lay Gap: medical VLMs achieve strong expert captioning but suffer significant lay-register degradation, while general-purpose VLMs produce more accessible language yet lack clinical precision, confirming that neither current paradigm adequately serves patient-facing communication.

Problem

• Medical Vision-Language Models
• Expert-Lay Gap
• Patient Accessibility
• Medical Lay Language Generation
• Clinical Communication

Innovation

• Medical Vision-Language Models
• Lay Language Generation
• UMLS Ontology
• HOVER Pipeline
• MedLayEval