Deep Learning in Diabetic Retinopathy Screening

The field of diabetic retinopathy screening is witnessing significant advancements with the application of deep learning techniques. Researchers are focusing on developing more accurate and reliable models for early detection and diagnosis of the disease. A key direction in this area is the development of semi-supervised learning approaches, which can effectively utilize both labeled and unlabeled data to improve model performance. Another important trend is the incorporation of explainability and interpretability in deep learning models, enabling clinicians to understand the decision-making process and build trust in the models. Noteworthy papers in this area include:

  • The survey on deep learning research in diabetic retinopathy screening, which provides a comprehensive overview of the field and outlines a practical agenda for future research.
  • The introduction of CalibrateMix, a targeted mixup-based approach for improving the calibration of semi-supervised learning models, which has shown promising results in achieving lower expected calibration error and superior accuracy.
  • The development of a semi-supervised multi-task learning framework for interpretable quality assessment of fundus images, which has demonstrated improved performance in overall quality assessment and provided interpretable feedback on capture conditions.

Sources

From Retinal Pixels to Patients: Evolution of Deep Learning Research in Diabetic Retinopathy Screening

CalibrateMix: Guided-Mixup Calibration of Image Semi-Supervised Models

Semi-Supervised Multi-Task Learning for Interpretable Quality As- sessment of Fundus Images

Stage Aware Diagnosis of Diabetic Retinopathy via Ordinal Regression

A Hybrid CNN-ViT-GNN Framework with GAN-Based Augmentation for Intelligent Weed Detection in Precision Agriculture

Explainable AI for Diabetic Retinopathy Detection Using Deep Learning with Attention Mechanisms and Fuzzy Logic-Based Interpretability

Built with on top of