Learning from spatially inhomogenous data: resolution-adaptive convolutions for multiple sclerosis lesion segmentation
Ivan Diaz, Florin Scherer, Yanik Berli, Roland Wiest, Helly Hammer, Robert Hoepner, Alejandro Leon Betancourt, Piotr Radojewski, Richard McKinley · 2025
Read Learning from spatially inhomogenous data: resolution-adaptive convolutions for multiple sclerosis lesion segmentation by Ivan Diaz, Florin Scherer, Yanik Berli, Roland Wiest, Helly Hammer, Robert Hoepner, Alejandro Leon Betancourt, Piotr Radojewski, Richard McKinley (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-2503-21829.txt.
Description
Ivan Diaz and eight co-authors build a cut net that learns straight from MRI scans took at grades that do not match. Care scans shift by who made the gear, by site, and by run type. The stock fix scales all scans to one grid. That fix brings in flaws from fills and drops fine points. Their net aims at cuts of sores from a brain and cord ill.
The build sets its core shapes with round wave terms at a fixed real size, through the e3nn frame. Cores can then be scaled to each scan's cell size, so no scan scale step is need. Trained on data from more than one site and from its own site, it beat stock U-Nets. That held in most test cases through both train plans. It also held for grades not seen in train.
Catalog record
| CREATOR | Ivan Diaz, Florin Scherer, Yanik Berli, Roland Wiest, Helly Hammer, Robert Hoepner, Alejandro Leon Betancourt, Piotr Radojewski, Richard McKinley |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 23 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-21829