MedFrameQA: A Multi-Image Medical VQA Benchmark for Clinical Reasoning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yu, Suhao, Wang, Haojin, Wu, Juncheng, Luo, Luyang, Wang, Jingshen, Xie, Cihang, Rajpurkar, Pranav, Yang, Carl, Yang, Yang, Wang, Kang, Yu, Yannan, Zhou, Yuyin
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918320214638592
author Yu, Suhao
Wang, Haojin
Wu, Juncheng
Luo, Luyang
Wang, Jingshen
Xie, Cihang
Rajpurkar, Pranav
Yang, Carl
Yang, Yang
Wang, Kang
Yu, Yannan
Zhou, Yuyin
author_facet Yu, Suhao
Wang, Haojin
Wu, Juncheng
Luo, Luyang
Wang, Jingshen
Xie, Cihang
Rajpurkar, Pranav
Yang, Carl
Yang, Yang
Wang, Kang
Yu, Yannan
Zhou, Yuyin
contents Real-world clinical practice demands multi-image comparative reasoning, yet current medical benchmarks remain limited to single-frame interpretation. We present MedFrameQA, the first benchmark explicitly designed to test multi-image medical VQA through educationally-validated diagnostic sequences. To construct this dataset, we develop a scalable pipeline that leverages narrative transcripts from medical education videos to align visual frames with textual concepts, automatically producing 2,851 high-quality multi-image VQA pairs with explicit, transcript-grounded reasoning chains. Our evaluation of 11 advanced MLLMs (including reasoning models) exposes severe deficiencies in multi-image synthesis, where accuracies mostly fall below 50% and exhibit instability across varying image counts. Error analysis demonstrates that models often treat images as isolated instances, failing to track pathological progression or cross-reference anatomical shifts. MedFrameQA provides a rigorous standard for evaluating the next generation of MLLMs in handling complex, temporally grounded medical narratives.
format Preprint
id arxiv_https___arxiv_org_abs_2505_16964
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MedFrameQA: A Multi-Image Medical VQA Benchmark for Clinical Reasoning
Yu, Suhao
Wang, Haojin
Wu, Juncheng
Luo, Luyang
Wang, Jingshen
Xie, Cihang
Rajpurkar, Pranav
Yang, Carl
Yang, Yang
Wang, Kang
Yu, Yannan
Zhou, Yuyin
Computer Vision and Pattern Recognition
Computation and Language
Real-world clinical practice demands multi-image comparative reasoning, yet current medical benchmarks remain limited to single-frame interpretation. We present MedFrameQA, the first benchmark explicitly designed to test multi-image medical VQA through educationally-validated diagnostic sequences. To construct this dataset, we develop a scalable pipeline that leverages narrative transcripts from medical education videos to align visual frames with textual concepts, automatically producing 2,851 high-quality multi-image VQA pairs with explicit, transcript-grounded reasoning chains. Our evaluation of 11 advanced MLLMs (including reasoning models) exposes severe deficiencies in multi-image synthesis, where accuracies mostly fall below 50% and exhibit instability across varying image counts. Error analysis demonstrates that models often treat images as isolated instances, failing to track pathological progression or cross-reference anatomical shifts. MedFrameQA provides a rigorous standard for evaluating the next generation of MLLMs in handling complex, temporally grounded medical narratives.
title MedFrameQA: A Multi-Image Medical VQA Benchmark for Clinical Reasoning
topic Computer Vision and Pattern Recognition
Computation and Language
url https://arxiv.org/abs/2505.16964