Anatomy-Preserving Latent Diffusion for Generation of Brain Segmentation Masks with Ischemic Infarct
1 st Lucia Borrego, 2 nd Vajira Thambawita, 3 rd Marco Ciuffreda, 4 th Inés del Val, 5 th Alejandro Dominguez, 6 th Josep Munuera · 2026
Read Anatomy-Preserving Latent Diffusion for Generation of Brain Segmentation Masks with Ischemic Infarct by 1 st Lucia Borrego, 2 nd Vajira Thambawita, 3 rd Marco Ciuffreda, 4 th Inés del Val, 5 th Alejandro Dominguez, 6 th Josep Munuera (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-2602-10167.txt.
Description
Good brain segmentation masks are scarce, and hand-labeling non-contrast CT scans is costly and inconsistent. A 2026 paper led by Lucia Borrego, with five co-authors, proposes an anatomy-preserving generative framework for synthesizing multi-class brain masks, including ischemic infarcts. A variational autoencoder trained only on masks learns an anatomical latent space, and a diffusion model in that space generates new samples from pure noise.
At inference, denoised latents decode through the frozen decoder, with optional rough control over lesion presence through a binary prompt. Generated masks preserve global anatomy, discrete tissue meanings, and realistic variability, avoiding artifacts common in pixel-space models. The framework offers a simple, scalable route to anatomy-aware masks for data-scarce imaging.
Catalog record
| CREATOR | 1 st Lucia Borrego, 2 nd Vajira Thambawita, 3 rd Marco Ciuffreda, 4 th Inés del Val, 5 th Alejandro Dominguez, 6 th Josep Munuera |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 26 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-10167