AutoRad-Lung combines radiomics with advanced vision-language models to improve lung nodule malignancy prediction, addressing the critical challenge of differentiating uncertain cases with similar visual characteristics.

• Integrates quantitative radiomic features with deep learning to enhance diagnostic accuracy
• Uses a guided prompting approach that mimics radiologist diagnostic workflows
• Demonstrates superior performance in uncertain cases where traditional methods struggle
• Provides explainable results that align with clinical decision-making processes

This innovation matters for medical practice by potentially improving early lung cancer detection rates, reducing false positives/negatives, and offering radiologists an AI assistant that works within their established diagnostic frameworks.

AutoRad-Lung: A Radiomic-Guided Prompting Autoregressive Vision-Language Model for Lung Nodule Malignancy Prediction