Abstract
Multimodalmedicalimagesprovidecomplementaryinformation for clinical diagnosis and treatment planning, but complete acquisitionsareofteninfeasiblebecauseofscanningtime,patientdiscomfort, contraindications, and cost. Medical image synthesis can recover missing modalities from available scans, and diffusion models have recently shown promising performance for this task. However, existing methods commonlyuseheterogeneouspreprocessingpipelines,architectures,and training protocols, hindering reproducibility and controlled comparison. WeproposennDiffusion,astandardized3Ddiffusionframeworkformedical image synthesis built upon the nnU-Net pipeline. It implements Denoising Diffusion Probabilistic Models, Flow Matching, and Diffusion Bridges using a common preprocessing pipeline, an automatically configured backbone, and a common training protocol. By inheriting the mature engineering design of nnU-Net, nnDiffusion enables systematic, reproducible, and practical evaluation of diffusion-based synthesis methods for 3D medical images. We validate nnDiffusion on the BraTS dataset with four MRI modalities: T1W, T2W, FLAIR, and T1CE, under both one-to-one and three-to-one synthesis settings. Experimental results demonstrate the feasibility, flexibility, and versatility of the proposed framework across diverse modality synthesis tasks. The code is available at https://github.com/PangHaowen-hub/nnDiffusion.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/SASHIMI_005.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=Z7IqvC4GtR
BibTex
@InProceedings{PanHao_nnDiffusion_MICCAISAT2026,
author = { Pang, Haowen AND Zhu, Pengli AND Chen, Shannan AND Hong, Xiaoming AND Ye, Chuyang},
title = { { nnDiffusion: A Standardized 3D Diffusion Framework for Medical Image Synthesis } },
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}
}