iLENS Predicts Alzheimer's Progression with Interpretable LLM Guidance

This paper introduces iLENS, a novel framework designed to improve the prediction of Alzheimer's Disease (AD) progression, particularly during its prodromal stage. AD is a complex neurodegenerative disorder, and accurate early prediction is vital for both understanding the disease and patient care. While traditional survival models are used for AD risk prediction, they often lack interpretability and the ability to incorporate natural language reasoning. iLENS addresses these limitations by employing a Mixture-of-Experts (MoE) architecture guided by a Large Language Model (LLM). The LLM's role is to synthesize both structured neuroimaging measurements and unstructured clinical information, using this combined understanding to direct the routing decisions of the expert models. This guidance allows iLENS to not only achieve strong predictive performance but also to subtype patients effectively. A key advantage of iLENS is its interpretability. The framework provides transparent, biologically sound explanations for its routing decisions, bridging the gap between high-performance survival analysis and actionable clinical decision support.