Offline Reinforcement Learning for Adaptive Treatment Strategies using Schrödinger Bridge Treatment Stitching

Imagine treating a chronic disease as a long road trip with many turns. Doctors have lots of historical GPS traces (patient histories) of trips that went well and badly, but they’re not allowed to experiment freely on real patients. This work designs a smarter GPS that learns from those old traces only, and then “stitches together” the best pieces of different trips into a new, better route using a sophisticated mathematical bridge. The goal is to recommend safer, more effective step‑by‑step treatment plans without doing risky trial‑and‑error on real people.