Nodule-Aligned Latent Space Learning with LLM-Driven Multimodal Diffusion for Lung Nodule Progression Prediction

TL;DR

This paper introduces NAMD, a novel multimodal diffusion framework that predicts lung nodule progression by generating follow-up images conditioned on baseline scans and patient data, improving early diagnosis accuracy.

Contribution

NAMD is the first to use a nodule-aligned latent space combined with LLM-driven control for lung nodule progression prediction from multimodal data.

Findings

Achieves AUROC of 0.805 for malignancy prediction

Synthesizes follow-up images closely matching real scans

Outperforms baseline and state-of-the-art methods

Abstract

Early diagnosis of lung cancer is challenging due to biological uncertainty and the limited understanding of the biological mechanisms driving nodule progression. To address this, we propose Nodule-Aligned Multimodal (Latent) Diffusion (NAMD), a novel framework that predicts lung nodule progression by generating 1-year follow-up nodule computed tomography images with baseline scans and the patient's and nodule's Electronic Health Record (EHR). NAMD introduces a nodule-aligned latent space, where distances between latents directly correspond to changes in nodule attributes, and utilizes an LLM-driven control mechanism to condition the diffusion backbone on patient data. On the National Lung Screening Trial (NLST) dataset, our method synthesizes follow-up nodule images that achieve an AUROC of 0.805 and an AUPRC of 0.346 for lung nodule malignancy prediction, significantly outperforming both baseline scans and state-of-the-art synthesis methods, while closely approaching the performance of real follow-up scans (AUROC: 0.819, AUPRC: 0.393). These results demonstrate that NAMD captures clinically relevant features of lung nodule progression, facilitating earlier and more accurate diagnosis.