Boundary-enhanced sparse transformer for generalizable and accurate medical image segmentation
Chaofan Li, Qiong Liu, Jianxiang Song · 2025
Read Boundary-enhanced sparse transformer for generalizable and accurate medical image segmentation by Chaofan Li, Qiong Liu, Jianxiang Song (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/pmc-12815953.txt.
Description
Chaofan Li, Qiong Liu, and Jianxiang Song present a 2025 paper on outlining organs and lesions in medical scans. Transformer networks lead this task, but their global attention wastes effort on irrelevant links and their decoders blur fine borders. Both flaws hurt diagnosis and treatment planning.
Their design attacks both problems. In the encoder, a frequency-based similarity score builds a small dictionary of key meanings, so attention stays sparse and relevant. In the decoder, a learned gradient operator biases attention toward object edges, restoring crisp contours. On a heart scan benchmark, accuracy rose and border error fell sharply against the second-best rival. On a skin lesion set it also led, while using far fewer parameters than typical transformers.
Catalog record
| CREATOR | Chaofan Li, Qiong Liu, Jianxiang Song |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 55 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |