LENS: A mammography-specific hybrid CNN-Transformer with lesion-aware evidence modeling
Duc Quy Hoang, Van Kien Cao, Tan Nhu Nguyen, Ngoc Son Nguyen · 2026
Read LENS: A mammography-specific hybrid CNN-Transformer with lesion-aware evidence modeling by Duc Quy Hoang, Van Kien Cao, Tan Nhu Nguyen, Ngoc Son Nguyen (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-0350720.txt.
Description
Duc Quy Hoang and co-authors build a mammography reader that forces itself to look at suspicious spots, not just the whole breast image. AI readers can cheat by leaning on background patterns, which makes their answers hard to trust. Their hybrid system blends a compact local-detail network with a transformer that scans wider context, plus a branch that gathers evidence from the regions most likely to hold a lesion.
On a large public mammography collection, the full version topped strong rival systems on accuracy and balanced scores while using fewer resources. Its chosen regions often overlapped the lesions human experts had marked. Hoang and colleagues argue that showing its evidence makes the system easier for radiologists to check, a key step toward clinic use.
Catalog record
| CREATOR | Duc Quy Hoang, Van Kien Cao, Tan Nhu Nguyen, Ngoc Son Nguyen |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 56420 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |