Deployable real-time spinal endoscopic instance segmentation with lightweight multi-scale attention mechanism
Qi Lai, Qiang Cai, JunYan Li, Chi Man Vong, Lei Jin, Lei Wang, Tao Yan, XiaoKun Liang · 2026
Read Deployable real-time spinal endoscopic instance segmentation with lightweight multi-scale attention mechanism by Qi Lai, Qiang Cai, JunYan Li, Chi Man Vong, Lei Jin, Lei Wang, Tao Yan, XiaoKun Liang (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/PMC13601576.txt.
Description
Qi Lai and colleagues introduce EndoSeg-RT in 2026, a lightweight AI system that outlines anatomy live during spinal endoscopy. Surgeons working through an endoscope must spot and protect nerves, bone, fat, and the ligamentum flavum in a narrow, glary view clouded by smoke and blood. Software that labels each structure in real time could guide the operation, but it must stay fast and steady even one frame at a time.
The design rebuilds the network's backbone, middle, and head around multi-scale attention that trains rich and runs lean. The team also released a clinically reviewed dataset of 610 images from 61 patients with per-structure labels. Their system matched or beat heavier rivals while using only 1.8 million parameters and modest computing power. It also generalized to a public dental imaging benchmark. Code and data are openly available, which the authors offer as a step toward deployable surgical assistance.
Catalog record
| CREATOR | Qi Lai, Qiang Cai, JunYan Li, Chi Man Vong, Lei Jin, Lei Wang, Tao Yan, XiaoKun Liang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 58073 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |