Mamba Goes HoME: Hierarchical Soft Mixture-of-Experts for 3D Medical Image Segmentation
Szymon Płotka, Maciej Chrabaszcz, Gizem Mert, Ewa Szczurek, Arkadiusz Sitek · 2025
Read Mamba Goes HoME: Hierarchical Soft Mixture-of-Experts for 3D Medical Image Segmentation by Szymon Płotka, Maciej Chrabaszcz, Gizem Mert, Ewa Szczurek, Arkadiusz Sitek (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-2507-06363.txt.
Description
Szymon Plotka and co-authors present HoME, a new design for splitting chest and body scans into parts with a neural net. Their paper appeared in twenty twenty-five. Three-dimension scans are long and vary a lot by scan type and quality. Long-range links in the scan still need to be caught at low cost.
HoME routes scan pieces to expert units in two steps. First, local groups of pieces go to experts tuned for close traits. Next, a global step mixes the group results to add wide context. The base is a fast state-space model called Mamba. Tests across three main scan types beat top past scores. The code is shared in full.
Catalog record
| CREATOR | Szymon Płotka, Maciej Chrabaszcz, Gizem Mert, Ewa Szczurek, Arkadiusz Sitek |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 88 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-06363