A Hybrid Mamba-SAM Architecture for Efficient 3D Medical Image Segmentation
Mohammadreza Gholipour Shahraki 1* Mehdi Rezaeian 2 Mohammad Ghasemzadeh 2,* · 2026
Read A Hybrid Mamba-SAM Architecture for Efficient 3D Medical Image Segmentation by Mohammadreza Gholipour Shahraki 1* Mehdi Rezaeian 2 Mohammad Ghasemzadeh 2,* (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-2602-00650.txt.
Description
Mohammadreza Gholipour Shahraki and colleagues adapt the Segment Anything Model to 3D medical segmentation, where domain shift, a 2D design, and costly fine-tuning limit it. Their Mamba-SAM hybrid keeps the SAM encoder frozen and adds Mamba modelling for long-range, linear-time context. Two variants are tested: a dual-branch design fusing frozen SAM features with a trainable VMamba encoder, and an adapter injecting volumetric Tri-Plane Mamba modules.
Catalog record
| CREATOR | Mohammadreza Gholipour Shahraki 1* Mehdi Rezaeian 2 Mohammad Ghasemzadeh 2,* |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36740 |
| LICENSE | http://creativecommons.org/publicdomain/zero/1.0/ |
| RIGHTS | CC0 |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602.00650