PuSH - Publication Server of Helmholtz Zentrum München

Export: TXT Text RIS Endnote (RIS) BIB BibTeX
Durrer, A.* ; Bieder, F.* ; Friedrich, P.* ; Menze, B.* ; Cattin, P.C.* ; Kofler, F.

fastWDM3D: Fast and accurate 3D healthy tissue inpainting.

In: (Deep Generative Models). Berlin [u.a.]: Springer, 2026. 171-181 (Lect. Notes Comput. Sc. ; 16128 LNCS)
DOI
Healthy tissue inpainting has many applications, for instance, generating pseudo-healthy baselines for tumor growth models or simplifying image registration. In prior editions of the BraTS Local Synthesis of Healthy Brain Tissue via Inpainting Challenge, denoising diffusion probabilistic models (DDPMs) demonstrated qualitatively convincing results but suffered from low sampling speed. To mitigate this limitation, we present a modified 3D wavelet diffusion model (WDM3D), denoted as fastWDM3D. Our proposed model employs a variance-preserving noise schedule and reconstruction losses over the full image as well as over the masked area only. Using fastWDM3D with only two time steps we achieved a SSIM of 0.8571, a MSE of 0.0079, and a PSNR of 22.26 on the BraTS inpainting test set. The 3D inpainting process took only 1.81 s per image. Compared to other DDPMs used for healthy brain tissue inpainting, our model is up to ∼ 800 × faster but still achieves superior performance metrics. Our proposed method, fastWDM3D, represents a promising approach for fast and accurate healthy tissue inpainting. Our code is available at https://github.com/AliciaDurrer/fastWDM3D.
Altmetric
Tags
Annotations
Special Publikation
Hide on homepage

Edit extra information
Edit own tags
Private
Edit own annotation
Private
Hide on publication lists
on hompage
Mark as special
publikation
Publication type Article: Conference contribution
Keywords 3d Diffusion Model ; Efficient ; Healthy Tissue Inpainting
Language english
Publication Year 2026
HGF-reported in Year 2026
ISSN (print) / ISBN 0302-9743
e-ISSN 1611-3349
Conference Title Deep Generative Models
Journal Lecture Notes in Computer Science
Quellenangaben Volume: 16128 LNCS, Issue: , Pages: 171-181 Article Number: , Supplement: ,
Publisher Springer
Publishing Place Berlin [u.a.]
Institute(s) Helmholtz Artifical Intelligence Cooperation Unit (HAICU)
POF-Topic(s) 30205 - Bioengineering and Digital Health
Research field(s) Enabling and Novel Technologies
PSP Element(s) G-530001-001
DOI 10.1007/978-3-032-05472-2_17
Scopus ID 105018585818
Erfassungsdatum 2025-10-22
[➜Report correction]