DIMA: DIffusing Motion Artifacts for unsupervised correction in brain MRI images
Paolo Angella, Luca Balbi, Fabrizio Ferrando, Paolo Traverso, Rosario Varriale, Vito Paolo Pastore *, Matteo Santacesaria * · 2025
Read DIMA: DIffusing Motion Artifacts for unsupervised correction in brain MRI images by Paolo Angella, Luca Balbi, Fabrizio Ferrando, Paolo Traverso, Rosario Varriale, Vito Paolo Pastore *, Matteo Santacesaria * (2025) 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-2504-06767.txt.
Description
Paolo Angella and co-authors remove motion blur from brain scans without needing clean training pairs in a 2025 paper. Patient motion degrades MRI quality and can force repeat scans, yet paired sharp and blurred images are rare in clinics. Their framework, DIMA, learns the look of motion artifacts from unpaired blurred scans alone.
A diffusion model trained on those scans then paints realistic artifacts onto clean images, manufacturing the pairs a correction network needs. The method avoids k-space tricks and sequence details, so it adapts across scanners and protocols. Across several data sets and scan planes, correction matched supervised rivals and generalized better to real clinical data. The authors aim to cut repeat scans and sharpen diagnosis.
Catalog record
| CREATOR | Paolo Angella, Luca Balbi, Fabrizio Ferrando, Paolo Traverso, Rosario Varriale, Vito Paolo Pastore *, Matteo Santacesaria * |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 32 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-06767