WangLab at MEDIQA-M3G 2024: Multimodal Medical Answer Generation using Large Language Models

TL;DR

This paper presents two top-performing solutions for multilingual multimodal medical answer generation, utilizing large language models and image classification techniques, achieving first and second place in the MEDIQA2024 shared task.

Contribution

The paper introduces novel multimodal medical answer generation methods using large language models and CLIP-based image classification, achieving state-of-the-art results in a competitive shared task.

Findings

First place in the competition leaderboard

Second place in the competition leaderboard

Insights into multimodal medical question answering

Abstract

This paper outlines our submission to the MEDIQA2024 Multilingual and Multimodal Medical Answer Generation (M3G) shared task. We report results for two standalone solutions under the English category of the task, the first involving two consecutive API calls to the Claude 3 Opus API and the second involving training an image-disease label joint embedding in the style of CLIP for image classification. These two solutions scored 1st and 2nd place respectively on the competition leaderboard, substantially outperforming the next best solution. Additionally, we discuss insights gained from post-competition experiments. While the performance of these two solutions have significant room for improvement due to the difficulty of the shared task and the challenging nature of medical visual question answering in general, we identify the multi-stage LLM approach and the CLIP image classification approach as promising avenues for further investigation.