3D-LLDM: LABEL-GUIDED 3D LATENT DIFFUSION MODEL FOR IMPROVING HIGH-RESOLUTION SYNTHETIC MR IMAGING IN HEPATIC STRUCTURE SEGMENTATION
Kyeonghun Kim 1 , Jaehyeok Bae 2 , Youngung Han 1,3 , Joo Young Bae 3 , Seoyoung Ju 1,4 , Junsu Lim 1,4 , Gyeongmin Kim 5 , Nam-Joon Kim 3,† , Woo Kyoung Jeong 6 , Ken Ying-Kai Liao 7 , Won Jae Lee 6,8 , Pa Hong 9 , and Hyuk-Jae Lee 3 · 2026
Read 3D-LLDM: LABEL-GUIDED 3D LATENT DIFFUSION MODEL FOR IMPROVING HIGH-RESOLUTION SYNTHETIC MR IMAGING IN HEPATIC STRUCTURE SEGMENTATION by Kyeonghun Kim 1 , Jaehyeok Bae 2 , Youngung Han 1,3 , Joo Young Bae 3 , Seoyoung Ju 1,4 , Junsu Lim 1,4 , Gyeongmin Kim 5 , Nam-Joon Kim 3,† , Woo Kyoung Jeong 6 , Ken Ying-Kai Liao 7 , Won Jae Lee 6,8 , Pa Hong 9 , and Hyuk-Jae Lee 3 (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-23845.txt.
Description
Kim and co-authors present 3D-LLDM, a model that makes synthetic liver MRI scans with matching masks. Labeled medical scans are scarce, which slows training of analysis tools. Their latent diffusion model is guided by labels for the liver, veins, and tumors. It trains on 720 real clinical scans from Samsung Medical Center. Its image quality score beats GANs by 70.9 percent and diffusion baselines by 26.7 percent. Added synthetic scans lift tumor segmentation scores across five network designs.
Catalog record
| CREATOR | Kyeonghun Kim 1 , Jaehyeok Bae 2 , Youngung Han 1,3 , Joo Young Bae 3 , Seoyoung Ju 1,4 , Junsu Lim 1,4 , Gyeongmin Kim 5 , Nam-Joon Kim 3,† , Woo Kyoung Jeong 6 , Ken Ying-Kai Liao 7 , Won Jae Lee 6,8 , Pa Hong 9 , and Hyuk-Jae Lee 3 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 20783 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.23845