MRI-to-sCT with Conditional Rectified Flow

Radiotherapy typically relies on MRI for tumor visualization and CT for dose calculation, but acquiring both scans introduces registration errors and slows clinical workflow. We develop sCTFlow, a conditional rectified flow generative model that synthesizes high-fidelity CT images directly from MRI. This MR-only approach aims to streamline treatment planning by producing anatomically consistent, dose-accurate synthetic CTs.