Learning Discriminative Representations for Fine-Grained Diabetic Retinopathy Grading

TL;DR

This paper proposes a fine-grained classification approach using a bilinear model and ordinal regression to improve diabetic retinopathy grading accuracy, leveraging discriminative features and ordinal information.

Contribution

It introduces a bilinear model combined with ordinal regression and metric loss for more accurate DR grading, addressing subtle class differences.

Findings

Superior performance on IDRiD dataset

Effective identification of discriminative fundus image regions

Improved grading accuracy over baseline methods

Abstract

Diabetic retinopathy (DR) is one of the leading causes of blindness. However, no specific symptoms of early DR lead to a delayed diagnosis, which results in disease progression in patients. To determine the disease severity levels, ophthalmologists need to focus on the discriminative parts of the fundus images. In recent years, deep learning has achieved great success in medical image analysis. However, most works directly employ algorithms based on convolutional neural networks (CNNs), which ignore the fact that the difference among classes is subtle and gradual. Hence, we consider automatic image grading of DR as a fine-grained classification task, and construct a bilinear model to identify the pathologically discriminative areas. In order to leverage the ordinal information among classes, we use an ordinal regression method to obtain the soft labels. In addition, other than only using a categorical loss to train our network, we also introduce the metric loss to learn a more discriminative feature space. Experimental results demonstrate the superior performance of the proposed method on two public IDRiD and DeepDR datasets.