Deep learning segmentation with uncertainty quantification for spinal tuberculosis on fat-suppressed T2-weighted MRI
Xingyu Duan, Jiaxing Wang, Linan Wang, Yichao Fan, Jiong Wang, Qianpeng Ma, Ningkui Niu · 2026
Read Deep learning segmentation with uncertainty quantification for spinal tuberculosis on fat-suppressed T2-weighted MRI by Xingyu Duan, Jiaxing Wang, Linan Wang, Yichao Fan, Jiong Wang, Qianpeng Ma, Ningkui Niu (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/PMC13461315.txt.
Description
Xingyu Duan and co-authors built in 2026 a deep learning system that outlines spinal tuberculosis lesions on MRI scans and, crucially, reports how unsure it is about each outline. Fuzzy lesion borders make hand tracing subjective, and black-box models leave doctors guessing where to distrust them. Scans from 210 patients across three centers trained and tested the system, which added a boundary-focused loss and repeated sampling to estimate uncertainty.
On the outside test set the model beat three established rivals on overlap accuracy, and its uncertainty tracked actual pixel errors closely. A threshold flagged unreliable regions with good sensitivity. Nine physicians reviewing cases with uncertainty overlays trusted the output more and finished faster, with trainees gaining most, and obvious failures were caught in most instances.
Catalog record
| CREATOR | Xingyu Duan, Jiaxing Wang, Linan Wang, Yichao Fan, Jiong Wang, Qianpeng Ma, Ningkui Niu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 45 KB |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |