Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Diabetic Retinopathy Grading with CLIP-based Ranking-Aware Adaptation:A Comparative Study on Fundus Image

Sungjun Cho

Abstract

Diabetic retinopathy (DR) is a leading cause of preventable blindness, and automated fundus image grading can play an important role in large-scale screening. In this work, we investigate three CLIP-based approaches for five-class DR severity grading: (1) a zero-shot baseline using prompt engineering, (2) a hybrid FCN-CLIP model augmented with CBAM attention, and (3) a ranking-aware prompting model that encodes the ordinal structure of DR progression. We train and evaluate on a combined dataset of APTOS 2019 and Messidor-2 (n=5,406), addressing class imbalance through resampling and class-specific optimal thresholding. Our experiments show that the ranking-aware model achieves the highest overall accuracy (93.42%, AUROC 0.9845) and strong recall on clinically critical severe cases, while the hybrid FCN-CLIP model (92.49%, AUROC 0.99) excels at detecting proliferative DR. Both substantially outperform the zero-shot baseline (55.17%, AUROC 0.75). We analyze the complementary strengths of each approach and discuss their practical implications for screening contexts.

Diabetic Retinopathy Grading with CLIP-based Ranking-Aware Adaptation:A Comparative Study on Fundus Image

Abstract

Diabetic retinopathy (DR) is a leading cause of preventable blindness, and automated fundus image grading can play an important role in large-scale screening. In this work, we investigate three CLIP-based approaches for five-class DR severity grading: (1) a zero-shot baseline using prompt engineering, (2) a hybrid FCN-CLIP model augmented with CBAM attention, and (3) a ranking-aware prompting model that encodes the ordinal structure of DR progression. We train and evaluate on a combined dataset of APTOS 2019 and Messidor-2 (n=5,406), addressing class imbalance through resampling and class-specific optimal thresholding. Our experiments show that the ranking-aware model achieves the highest overall accuracy (93.42%, AUROC 0.9845) and strong recall on clinically critical severe cases, while the hybrid FCN-CLIP model (92.49%, AUROC 0.99) excels at detecting proliferative DR. Both substantially outperform the zero-shot baseline (55.17%, AUROC 0.75). We analyze the complementary strengths of each approach and discuss their practical implications for screening contexts.
Paper Structure (31 sections, 8 figures, 5 tables)

This paper contains 31 sections, 8 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Deep learning for DR grading.
  4. Vision-language models in medical imaging.
  5. Ordinal classification and ranking-aware learning.
  6. Attention mechanisms for lesion localization.
  7. Methods
  8. Zero-shot CLIP Baseline
  9. Hybrid FCN-CLIP with CBAM Attention
  10. Ranking-aware Prompting Model
  11. Dataset and Preprocessing
  12. Experimental Setup
  13. Results
  14. Overall Performance
  15. Error Distribution
  16. ...and 16 more sections

Figures (8)

  • Figure 1: Attention map comparison for a mild DR case. Left: zero-shot CLIP (diffuse, anatomy-focused). Right: Hybrid FCN-CLIP with CBAM (concentrated on lesion regions).
  • Figure 2: Training and validation curves: Ranking-aware prompting model.
  • Figure 3: Training and validation curves: Hybrid FCN-CLIP model.
  • Figure 4: Confusion matrix: Ranking-aware prompting model.
  • Figure 5: One-vs-rest ROC curves: Ranking-aware model.
  • ...and 3 more figures