Traj-Evolve: A Self-Evolving Multi-Agent System for Patient Trajectory Modeling in Lung Cancer Early Detection

2026-06-01

AI Summary

This study addresses the limitations of existing methods in effectively leveraging historical case experience and handling sparse, noisy, and long-sequence multimodal data when modeling longitudinal electronic health records. To overcome these challenges, the authors propose a self-evolving multi-agent system that integrates a non-parametric experience buffer with a dual evolutionary mechanism based on multi-agent reinforcement learning, enabling retrieval-augmented early prediction of lung cancer. The work introduces a novel leave-one-out cross-retrieval strategy to unify training and inference, and further incorporates rejection-sampling-based trajectory indexing, memory retrieval, and reward-ranking fine-tuning to achieve, for the first time, co-evolution between agents and the memory bank. Evaluated on over five years of multimodal electronic health records, the model significantly outperforms nine strong baselines across the general population and, notably, within the challenging subgroup of never-smokers.

Abstract

Modeling patient trajectories from longitudinal electronic health records (EHRs) requires reasoning over sparse, noisy, and long-context multimodal sequences. Existing LLM-based multi-agent systems address context length but process patients in isolation, failing to mirror how clinicians leverage accumulated experience from similar prior cases. We present Traj-Evolve, a self-evolving multi-agent system with two complementary evolving mechanisms. First, an Experience Pool (ExPool) acts as a non-parametric memory, indexing rejection-sampled reasoning traces to retrieve similar patients as few-shot contexts. Second, multi-agent reinforcement learning (MARL) via reward-ranked fine-tuning parametrically optimizes inter-agent and agent-memory collaboration. A leave-one-out cross-retrieval strategy unifies the two, aligning training- and inference-time behavior under retrieval augmentation. On a lung cancer prediction task utilizing up to five years of multimodal EHRs, Traj-Evolve outperforms 9 strong baselines on the overall population and a challenging never-smoker population. Analysis of the evolving dynamics highlights three key findings: (1) expanding the ExPool shifts optimal retrieval from diverse to specific samples; (2) under MARL, the manager agent's prediction loss converges quickly while the worker agents' temporal reasoning continues to benefit from more verified patients; and (3) the two mechanisms are complementary on the predicted risk, where ExPool improves specificity while MARL improves sensitivity.

Problem

• patient trajectory modeling
• lung cancer early detection
• electronic health records
• multimodal sequences
• clinical decision support

Innovation

• self-evolving multi-agent system
• experience pool
• multi-agent reinforcement learning
• retrieval-augmented reasoning
• patient trajectory modeling