Dual cross-attentive mutual teaching for semi-supervised 3D medical segmentation
Weiping Ma · 2026
Read Dual cross-attentive mutual teaching for semi-supervised 3D medical segmentation by Weiping Ma (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/plos-pone-0352358.txt.
Description
Weiping Ma proposes a new way to outline organs in three-dimensional medical scans in this 2026 paper. Drawing those outlines by hand is slow expert work. Semi-supervised learning cuts the burden by learning from a few labelled scans plus many unlabelled ones. Ma's framework sets two networks to teach each other. A cross-attention module lets them exchange detailed features as they learn.
The two branches follow a teacher-student design built on a VNet backbone. Extra losses keep their feature distributions aligned and their classes cleanly separated. Knowledge flows in both directions, so each network corrects the other. Tests used left atrium scans and pancreatic scans, two hard segmentation targets. The method outperformed existing approaches on both. Hospitals could gain accurate scan outlines while labelling only a fraction of their image archives.
Catalog record
| CREATOR | Weiping Ma |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53842 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |