Query-Driven Retinal Layer Segmentation in OCT Using Cross-Attentive Feature Learning
Nebras Sobahi, Salih Taha Alperen Özçelik, Orhan Atila, Abdulkadir Sengur, Muhammed Halil Akpınar · 2026
Read Query-Driven Retinal Layer Segmentation in OCT Using Cross-Attentive Feature Learning by Nebras Sobahi, Salih Taha Alperen Özçelik, Orhan Atila, Abdulkadir Sengur, Muhammed Halil Akpınar (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-13256827.txt.
Description
Nebras Sobahi and co-authors present in 2026 RetiQueryNet, a deep learning model that segments the layers of the retina in optical coherence tomography scans. It uses learned query embeddings instead of purely pixel-wise prediction. Layer boundaries guide diagnosis of diseases such as macular degeneration and diabetic macular edema, and pixel methods often break anatomical consistency in faint or distorted regions.
Each retinal layer is represented by a query that attends to image features through cross-attention in a transformer-based design. The model reached a mean Dice score of 0.934 and beat U-Net, DeepLabV3, and other baselines, with the clearest gains in layers whose boundaries are hardest to trace. Query-based modeling may transfer to other medical segmentation tasks.
Catalog record
| CREATOR | Nebras Sobahi, Salih Taha Alperen Özçelik, Orhan Atila, Abdulkadir Sengur, Muhammed Halil Akpınar |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |