sCTFlow: 3D MRI-to-sCT with Conditional Rectified Flow

Pharuj Rajborirug, Supan Tungjitkusolmun, Sarun Gulyanon

Keywords: synthetic CT, MRI-only radiotherapy, rectified flow, flow matching, medical image synthesis

TL;DR: sCTFlow is a Conditional Rectified Flow model that produces realistic 3D synthetic CT from MRI. It also captures meaningful HU distributions, showing strong potential for MR-only radiotherapy workflows.

Abstract: MRI and CT images are both crucial for radiotherapy planning, since MRI provides superior soft-tissue contrast for tumor delineation, while CT provides Hounsfield units (HU) required for dose calculation. MR-only radiotherapy offers important advantages, including reduced registration errors, elimination of additional radiation exposure, and streamlined clinical workflows. Generating synthetic CT (sCT) from MRI remains challenging due to the need for realistic HU reconstruction and the high computational demands of processing large 3D image volumes. We propose sCTFlow, a Conditional Rectified Flow (CRF) framework for 3D MRI-to-sCT translation. Unlike diffusion probabilistic models (DDPM), sCTFlow learns a deterministic velocity field mapping noise to data, ensuring stability and requiring fewer sampling steps. Our architecture, a 3D Attention U-Net, conditions on MRI and organ segmentation via feature-wise linear modulation to predict velocity fields, which are subsequently converted into HU estimates. We evaluated our approach on the SynthRAD2023 dataset. sCTFlow achieves MAE of 82.51$\pm$18.96 HU, PSNR of 26.70, and SSIM of 0.824. We also investigated our method and found that the model captures HU distributions rather than relying on simple intensity transformations, indicating its capacity to model underlying CT characteristics. These findings demonstrate that sCTFlow has potential for reliable and clinically applicable MR-only radiotherapy workflows.

Primary Subject Area: Image Synthesis

Secondary Subject Area: Application: Radiology

Registration Requirement: Yes

Reproducibility: https://github.com/MaiRajborirug/CRFlow

Visa & Travel: Yes

Read CFP & Author Instructions: Yes

Originality Policy: Yes

Single-blind & Not Under Review Elsewhere: Yes

LLM Policy: Yes

Submission Number: 268