Exploring the Boundaries of GPT-4 in Radiology

TL;DR

This paper evaluates GPT-4's performance on radiology report tasks, showing it often surpasses or matches specialized models, with strong zero-shot results and potential for clinical application.

Contribution

It demonstrates GPT-4's competitive performance in radiology text tasks, highlighting its zero-shot capabilities and potential as a generalist model in medical NLP.

Findings

GPT-4 outperforms SOTA models in temporal sentence similarity and natural language inference.

GPT-4 matches supervised models in findings summarisation with example prompting.

Error analysis shows GPT-4 has sufficient radiology knowledge with occasional nuanced errors.

Abstract

The recent success of general-domain large language models (LLMs) has significantly changed the natural language processing paradigm towards a unified foundation model across domains and applications. In this paper, we focus on assessing the performance of GPT-4, the most capable LLM so far, on the text-based applications for radiology reports, comparing against state-of-the-art (SOTA) radiology-specific models. Exploring various prompting strategies, we evaluated GPT-4 on a diverse range of common radiology tasks and we found GPT-4 either outperforms or is on par with current SOTA radiology models. With zero-shot prompting, GPT-4 already obtains substantial gains ($\approx$ 10% absolute improvement) over radiology models in temporal sentence similarity classification (accuracy) and natural language inference ($F_1$). For tasks that require learning dataset-specific style or schema (e.g. findings summarisation), GPT-4 improves with example-based prompting and matches supervised SOTA. Our extensive error analysis with a board-certified radiologist shows GPT-4 has a sufficient level of radiology knowledge with only occasional errors in complex context that require nuanced domain knowledge. For findings summarisation, GPT-4 outputs are found to be overall comparable with existing manually-written impressions.