R2GenGPT: Radiology Report Generation with Frozen LLMs

TL;DR

R2GenGPT leverages frozen large language models with a visual alignment module to generate radiology reports efficiently, achieving state-of-the-art results with minimal additional training.

Contribution

The paper introduces R2GenGPT, a novel method that aligns visual features with LLMs using a lightweight module, enabling effective radiology report generation without fine-tuning the entire model.

Findings

Achieves state-of-the-art performance in R2Gen tasks.

Requires training only 5 million parameters, 0.07% of total.

Demonstrates high training efficiency and rapid convergence.

Abstract

Large Language Models (LLMs) have consistently showcased remarkable generalization capabilities when applied to various language tasks. Nonetheless, harnessing the full potential of LLMs for Radiology Report Generation (R2Gen) still presents a challenge, stemming from the inherent disparity in modality between LLMs and the R2Gen task. To bridge this gap effectively, we propose R2GenGPT, which is a novel solution that aligns visual features with the word embedding space of LLMs using an efficient visual alignment module. This innovative approach empowers the previously static LLM to seamlessly integrate and process image information, marking a step forward in optimizing R2Gen performance. R2GenGPT offers the following benefits. First, it attains state-of-the-art (SOTA) performance by training only the lightweight visual alignment module while freezing all the parameters of LLM. Second, it exhibits high training efficiency, as it requires the training of an exceptionally minimal number of parameters while achieving rapid convergence. By employing delta tuning, our model only trains 5M parameters (which constitute just 0.07\% of the total parameter count) to achieve performance close to the SOTA levels. Our code is available at https://github.com/wang-zhanyu/R2GenGPT.