Self-Supervised Representation Learning for Data-Efficient DRIL Classification in OCT Images
Pavithra Kodiyalbail Chakrapani, Akshat Tulsani, Preetham Kumar, Geetha Maiya, Sulatha Venkataraya Bhandary, Steven Fernandes · 2025
Read Self-Supervised Representation Learning for Data-Efficient DRIL Classification in OCT Images by Pavithra Kodiyalbail Chakrapani, Akshat Tulsani, Preetham Kumar, Geetha Maiya, Sulatha Venkataraya Bhandary, Steven Fernandes (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-12732198.txt.
Description
Chakrapani and five co-authors built an eye image classifier that learns from unlabeled scans in twenty twenty-five. Their target, DRIL, is a disorganization of the retina's inner layers seen in diabetic eye swelling. It strongly tracks vision quality, but expert-labeled training images are scarce.
Their two-stage method first learns from over one hundred eight thousand unlabeled retina scans using self-supervision, a setup where the model creates its own training signals. A space-aware variant of the BYOL method teaches it retinal anatomy. Fine-tuning then uses only eight hundred twenty-three labeled scans to classify DRIL. Accuracy reached over ninety-nine percent, beating models transferred from general image training by a wide margin. The authors show domain-specific self-supervision suits rare retinal diseases.
Catalog record
| CREATOR | Pavithra Kodiyalbail Chakrapani, Akshat Tulsani, Preetham Kumar, Geetha Maiya, Sulatha Venkataraya Bhandary, Steven Fernandes |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |