A Discordance-Aware Multimodal Framework with Multi-Agent Clinical Reasoning

ArXi:2604.16333v1 Announce Type: new Knee osteoarthritis frequently exhibits discordance between structural damage observed in imaging and patient-reported symptoms such as pain. This mismatch complicates clinical interpretation and patient stratification and remains insufficiently modeled in existing decision systems. We propose a discordance aware multimodal framework that combines machine learning prediction models with a tool grounded multi agent reasoning system.