Tag
MA-RAG is a multi-agent retrieval-augmented generation framework for query-driven summarization of longitudinal Parkinson's disease clinical assessments, achieving substantial improvements in factual precision and reducing hallucination rates compared to baseline methods.
This paper proposes a structural variant of the Factorial Hidden Markov Model to analyze and interpret disease trajectories in Type 2 diabetes patients using electronic health records, revealing clinically meaningful patterns and progression pathways.
This paper proposes a self-supervised Mamba-based model to learn effective representations from electronic health records for improved patient subtyping, demonstrating better performance than baseline models on real-world datasets.
This paper introduces LANTERN, a neural network framework for estimating health-state transition probabilities from irregular longitudinal data, with applications to long-term care insurance. It outperforms traditional methods in discrimination and calibration for severe disability and mortality prediction.
LifeSentence finetunes a 24B-parameter language model on structured natural-language records from a longitudinal panel study (SOEP), achieving superior prediction of life outcomes and enabling counterfactual queries about human biographies.
This review paper proposes a unified framework for intervention-aware disease trajectory modeling in clinical AI, addressing static prediction failures by incorporating treatment confounder feedback and informative observation patterns.