Traj-Evolve: A Self-Evolving Multi-Agent System for Patient Trajectory Modeling in Lung Cancer Early Detection
Researchers have developed Traj-Evolve, a novel multi-agent system designed to improve early lung cancer detection by modeling patient trajectories. This system utilizes an Experience Pool to retrieve similar past patient cases and employs multi-agent reinforcement learning to optimize collaboration between agents and memory. Experiments show Traj-Evolve outperforms existing methods, particularly in identifying risk among never-smokers, by enhancing both specificity and sensitivity through its evolving mechanisms. AI
IMPACT This system could enhance early disease detection by leveraging accumulated clinical experience, potentially improving patient outcomes.