Coarse-to-Fine Personalized LLM Impressions for Streamlined Radiology Reports

TL;DR

This paper introduces a coarse-to-fine LLM-based framework that automatically generates and personalizes radiology impressions, aiming to reduce radiologist workload while ensuring accuracy and style alignment.

Contribution

It presents a novel multi-stage approach using open-source LLMs fine-tuned on clinical data and reinforcement learning to personalize and improve radiology report impressions.

Findings

Reduces radiologist workload significantly

Achieves high factual accuracy in generated impressions

Personalizes impressions to individual radiologist styles

Abstract

The manual creation of the "Impression" section in radiology reports is a primary driver of radiologist burnout. To address this challenge, we propose a coarse-to-fine framework that leverages open-source large language models (LLMs) to automatically generate and personalize impressions from clinical findings. The system first produces a draft impression and then refines it using machine learning and reinforcement learning from human feedback (RLHF) to align with individual radiologists' styles while ensuring factual accuracy. We fine-tune LLaMA and Mistral models on a large dataset of reports from the University of Chicago Medicine. Our approach is designed to significantly reduce administrative workload and improve reporting efficiency while maintaining high standards of clinical precision.