CA2PNet: a context-aware multi-scale architecture with adaptive attention and progressive dilated convolutions for biomedical image segmentation
Aman Kumar Singh, Ashish Ranjan, Manas Ranjan Prusty, Nitish Katal · 2026
Read CA2PNet: a context-aware multi-scale architecture with adaptive attention and progressive dilated convolutions for biomedical image segmentation by Aman Kumar Singh, Ashish Ranjan, Manas Ranjan Prusty, Nitish Katal (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/pmc-13279314.txt.
Description
Aman Kumar Singh and colleagues published a 2026 paper presenting CA2PNet, a deep learning network that segments biomedical images, outlining structures such as lesions or organs pixel by pixel. Accurate outlines underpin diagnosis, surgery planning, and tumor measurement, yet anatomy varies widely between patients and target regions are often small. The design builds on two earlier networks and adds four components: a spatial attention module, global max pooling, enhanced pyramid pooling for features at several scales, and progressively widening convolutions.
On the Kvasir-SEG endoscopy dataset the network scored a mean overlap of 85.15 percent, and on the BUSI breast ultrasound dataset 82.78 percent, beating the state-of-the-art models they compared against. Tests supported the gains. The authors credit finer local detail kept alongside whole-image context, which helps the model hold boundaries across scales. They position the architecture as a general tool for computer-aided diagnosis pipelines.
Catalog record
| CREATOR | Aman Kumar Singh, Ashish Ranjan, Manas Ranjan Prusty, Nitish Katal |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 69 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |