What question did this study set out to answer?

This research aims to improve fovea segmentation accuracy in fundus images for diabetic retinopathy screening by leveraging anatomical landmarks.

February 26, 2026Open Access

Contextual anatomy-guided deep learning for accurate fovea segmentation in diabetic retinopathy fundus images

Key Points

This research aims to improve fovea segmentation accuracy in fundus images for diabetic retinopathy screening by leveraging anatomical landmarks.
Developed a new dataset (IDRiD-RETA-FV) with annotated anatomical structures for training and testing.
Implemented the MNv4Fovea framework using anatomical context for fovea segmentation.
Evaluated performance on the held-out test set using defined metrics such as IoU and F1 score.
Applied GEV-based augmentation to enhance detection rates compared to baseline methods.
Achieved a fovea IoU of 0.812 and F1 score of 0.894 on the test set.
The proposed approach had a detection rate of 98.4%, significantly outperforming the 59.0% rate from baseline methods.
Demonstrated competitive localization performance on external datasets (REFUGE, MESSIDOR) with state-of-the-art results.

Abstract

Abstract Accurate fovea segmentation in fundus images is a critical step in diabetic retinopathy screening; however, it remains a challenging task due to the indistinct boundaries of the fovea. Beyond simple localization, precise segmentation offers essential clinical value for Diabetic Macular Edema (DME) management, as treatment decisions–specifically the choice between intravitreal anti-VEGF injection for center-involved DME and laser therapy for extrafoveal edema–depend on the accurate delineation of the foveal region. While existing methods often rely on increasing model architecture complexity, the potential of anatomical context within the training process remains under-explored. This paper presents a data-centric approach that leverages contextual information to robustly identify the fovea. We demonstrate that progressively incorporating key anatomical landmarks–the optic disc, retina, and blood vessels–into training labels significantly enhances fovea detection. To facilitate this, we developed IDRiD-RETA-FV, a meticulously annotated dataset comprising 81 images (54 training, 27 testing) with complete anatomical structures (inter-observer F1=0.98), and introduce MNv4Fovea, a framework designed to explicitly exploit these anatomical inter-dependencies through a multi-class constraints mechanism. Evaluation on the held-out test set with verified ground truth demonstrates excellent segmentation performance (fovea IoU = 0.812, F1 = 0.894, AED = 4.06 pixels). To demonstrate the efficacy of our synthesis strategy, our GEV-based augmentation technique achieves a detection rate of 98.4% compared to 59.0% for baseline geometric augmentation (paired t-test: t = 8.536, p < 0.001, Cohen’s d = 1.093). Cross-dataset evaluation on REFUGE, MESSIDOR, and ARIA demonstrates competitive localization performance, achieving state-of-the-art Average Euclidean Distance on REFUGE (22.46 ± 18.73 pixels) and MESSIDOR (6.52 ± 5.89 pixels) with robust generalization across diverse imaging protocols. These results establish that explicit anatomical context, rather than mere model complexity, is key to accurate fovea segmentation, offering a robust paradigm for medical image analysis.

AI에게 질문

Bookmark

View Full Paper

Cite This Study

Chankhachon et al. (Tue,) studied this question.

synapsesocial.com/papers/699fe36b95ddcd3a253e7341 https://doi.org/https://doi.org/10.1038/s41598-026-40287-y

AI에게 질문

Bookmark

View Full Paper