Nested Radially Monotone Polar Occupancy Estimation: Clinically-Grounded Optic Disc and Cup Segmentation for Glaucoma Screening
2026-04-10 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors created a new deep learning method called NPS-Net to better segment the optic disc and cup in eye photos, which is important for glaucoma screening. Their method ensures the shapes follow specific clinical rules, making the results more reliable, especially when tested on different datasets without extra training. They tested it on seven datasets and found it to be very accurate and consistent, outperforming previous methods in important metrics like Dice score and anatomical validity. This means their approach can potentially improve automated glaucoma detection.
optic discoptic cupglaucoma screeningdeep learningsegmentationpolar coordinatesnested structureradially monotoneDice scorecross-dataset generalization
Authors
Rimsa Goperma, Rojan Basnet, Liang Zhao
Abstract
Valid segmentation of the optic disc (OD) and optic cup (OC) from fundus photographs is essential for glaucoma screening. Unfortunately, existing deep learning methods do not guarantee clinical validness including star-convexity and nested structure of OD and OC, resulting corruption in diagnostic metric, especially under cross-dataset domain shift. To adress this issue, this paper proposed NPS-Net (Nested Polar Shape Network), the first framework that formulates the OD/OC segmentation as nested radially monotone polar occupancy estimation.This output representation can guarantee the aforementioned clinical validness and achieve high accuracy. Evaluated across seven public datasets, NPS-Net shows strong zero-shot generalization. On RIM-ONE, it maintains 100% anatomical validity and improves Cup Dice by 12.8% absolute over the best baseline, reducing vCDR MAE by over 56%. On PAPILA, it achieves Disc Dice of 0.9438 and Disc HD95 of 2.78 px, an 83% reduction over the best competing method.