Automated analysis of diabetic retinopathy using vessel segmentation maps as inductive bias

TL;DR

This paper proposes a novel automated method for diabetic retinopathy grading using vessel segmentation maps from OCTA images, combining synthetic data training with real data application, and achieves comparable performance to existing models.

Contribution

It introduces a new approach that integrates vessel segmentation maps as inductive bias in DR grading, utilizing synthetic OCTA images for training.

Findings

Method performs as well as baseline models on DR analysis challenge

Synthetic OCTA images effectively train vessel segmentation networks

Vessel segmentation maps enhance DR grading accuracy

Abstract

Recent studies suggest that early stages of diabetic retinopathy (DR) can be diagnosed by monitoring vascular changes in the deep vascular complex. In this work, we investigate a novel method for automated DR grading based on optical coherence tomography angiography (OCTA) images. Our work combines OCTA scans with their vessel segmentations, which then serve as inputs to task specific networks for lesion segmentation, image quality assessment and DR grading. For this, we generate synthetic OCTA images to train a segmentation network that can be directly applied on real OCTA data. We test our approach on MICCAI 2022's DR analysis challenge (DRAC). In our experiments, the proposed method performs equally well as the baseline model.