Argus: Benchmarking and Enhancing Vision-Language Models for 3D Radiology Report Generation

TL;DR

This paper introduces Argus, a new benchmark dataset and training methodology for vision-language models in 3D radiology report generation, significantly improving performance on CT scan reports.

Contribution

It creates the largest public 3D CT report dataset and develops a comprehensive training recipe, leading to the state-of-the-art Argus models for 3D radiology report generation.

Findings

Argus models outperform previous methods across various sizes and resolutions.

The benchmark enables robust evaluation of VLMs on 3D radiology data.

Optimal training strategies significantly enhance report generation quality.

Abstract

Automatic radiology report generation holds significant potential to streamline the labor-intensive process of report writing by radiologists, particularly for 3D radiographs such as CT scans. While CT scans are critical for clinical diagnostics, they remain less explored compared to 2D radiographs. To date, there has been no comprehensive benchmark for 3D radiograph report generation (3DRRG), nor sufficient investigation into the optimal training strategies for Vision Language Models (VLMs) in this context, particularly with respect to vision encoder choices, visual token compression, and model scaling. In this work, we make three key contributions. We curate **CT-3DRRG**, the largest **publicly** available 3D CT-report dataset, establishing a robust and diverse benchmark for evaluating VLM performance on 3DRRG. Furthermore, we propose a comprehensive training recipe for building high-performing VLMs for 3DRRG, exploring key factors such as vision encoder pretraining strategies, visual token compression, and the impact of data & model scale. Guided by these findings, we introduce **Argus**, a state-of-the-art family of VLMs that achieve superior performance across different model sizes and input 3D medical image resolutions, efficiently processing high-resolution 3D images up to $512 \times 512 \times 256$[^1].