MOTOR: Multimodal Optimal Transport via Grounded Retrieval in Medical Visual Question Answering

TL;DR

MOTOR introduces a multimodal retrieval and re-ranking method using grounded captions and optimal transport to improve medical visual question answering accuracy by providing more relevant clinical context.

Contribution

It presents a novel multimodal retrieval and re-ranking approach that leverages grounded captions and optimal transport for better context relevance in MedVQA.

Findings

Achieves 6.45% higher accuracy than state-of-the-art methods.

Outperforms existing retrieval approaches in clinical relevance.

Validated by empirical analysis and human expert evaluation.

Abstract

Medical visual question answering (MedVQA) plays a vital role in clinical decision-making by providing contextually rich answers to image-based queries. Although vision-language models (VLMs) are widely used for this task, they often generate factually incorrect answers. Retrieval-augmented generation addresses this challenge by providing information from external sources, but risks retrieving irrelevant context, which can degrade the reasoning capabilities of VLMs. Re-ranking retrievals, as introduced in existing approaches, enhances retrieval relevance by focusing on query-text alignment. However, these approaches neglect the visual or multimodal context, which is particularly crucial for medical diagnosis. We propose MOTOR, a novel multimodal retrieval and re-ranking approach that leverages grounded captions and optimal transport. It captures the underlying relationships between the query and the retrieved context based on textual and visual information. Consequently, our approach identifies more clinically relevant contexts to augment the VLM input. Empirical analysis and human expert evaluation demonstrate that MOTOR achieves higher accuracy on MedVQA datasets, outperforming state-of-the-art methods by an average of 6.45%. Code is available at https://github.com/BioMedIA-MBZUAI/MOTOR.