Abstract
Cone-beam CT (CBCT) is routinely acquired in radiotherapy but suffers from severe artifacts and unreliable Hounsfield Unit (HU) values, limiting its direct use for dose calculation. Synthetic CT (sCT)generationfromCBCTisthereforeanimportanttask,yetpaired CBCT–CTdataareoftenunavailableorunreliableduetotemporalgaps, anatomicalvariation,andregistrationerrors.Inthiswork,weintroduce Rectified Flow (RF) into unpaired CBCT-to-CT translation in medical imaging.AlthoughRFistheoreticallycompatiblewithunpairedlearning throughdistribution-levelcouplinganddeterministictransport,itspractical effectiveness under small medical datasets and limited batch sizes remains underexplored. Random or batch-local pseudo pairing can produce noisy transport targets when endpoint samples differ substantially inanatomy.Toaddressthischallenge,weproposeRetrieval-Augmented FlowMatching(RAFM),whichconstructsretrieval-guidedpseudopairs usingafrozenDINOv3encoderandarollingCTmemorybank.Theretrievedendpointsreducegrossstructuralmismatchwhiletherollingbank retainspairingvariationacrosstraining.ExperimentsonSynthRAD2023 underasubject-disjointunpairedprotocolshowthatRAFMoutperforms existing methods in terms of both image appearance and high-level semantic consistency. The code is available at https://github.com/HiLabgit/RAFM.git.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/SASHIMI_002.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=DK6KrbpRMQ
BibTex
@InProceedings{ZhoXia_RAFM_MICCAISAT2026,
author = { Zhou, Xianhao AND Wu, Jianghao AND Zhong, Lanfeng AND Zhao, Ku AND He, Jinlong AND Zhang, Shaoting AND Wang, Guotai},
title = { { RAFM: Retrieval-Augmented Flow Matching for Unpaired CBCT-to-CT Translation } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17258},
month = {pending},
page = {pending}
}