To improve the segmentation of retinal blood vessels using an AI-driven saliency-guided boundary refinement framework (SGB-Net) that addresses challenges such as low contrast, complex vessel geometry, and the degradation of segmentation performance due to pathological artifacts.
Key Findings:
Achieved Dice scores of 98.30%, 78.40%, and 84.60% on the DRIVE, STARE, and CHASE_DB1 datasets, respectively, indicating high accuracy in segmentation.
AUC values reached up to 0.9899, demonstrating excellent model performance.
Improved preservation of thin vessels and enhanced boundary continuity compared to existing methods.
Interpretation:
The SGB-Net effectively combines boundary refinement with multi-scale and attention-based feature learning. This combination makes it robust against noise and pathological variations, making it suitable for large-scale digital pathology applications.
Limitations:
Future work needed to improve sensitivity in detecting fine vessels.
Potential extension required for other medical imaging modalities to enhance applicability.
Conclusion:
The proposed framework significantly enhances retinal vessel segmentation performance, addressing key limitations of existing techniques.
An artificial intelligence–based optical coherence tomography pathway met noninferiority criteria for false-positive diabetic macular edema referrals and was associated with fewer referral decisions in a randomized clinical trial.
A nationwide Swedish cohort found a higher relative risk of anterior ischemic optic neuropathy among GLP-1 receptor agonist users, but the overall risks were low.
