To develop and validate an improved YOLOv11-based deep learning algorithm for accurate macular edema detection in optical coherence tomography (OCT) images, and to build DeepME, a lightweight system for diagnosis and treatment recommendations.
Approach:
Dataset Compilation: A comprehensive dataset was compiled from hospital clinical data and public OCT resources, covering macular edema and other retinal diseases.
External Validation: External validation was performed using an anti-VEGF cohort of 336 eyes from 300 patients with diabetic retinopathy or retinal vein occlusion.
Algorithm Improvement: The improved YOLOv11n integrated the Convolutional Block Attention Module (CBAM) to enhance feature extraction.
System Integration: DeepME combined the YOLOv11 detector with an optimized DeepSeek model, current clinical guidelines, and expert knowledge to create a cohesive diagnostic tool.
Key Findings:
DeepME achieved accuracy of 0.980, specificity of 0.990, sensitivity of 0.970, precision of 0.990, F1-score of 0.980, and AUC of 0.9993.
Grad-CAM visualizations confirmed precise localization of cystoid macular edema within anatomically correct retinal layers.
In the anti-VEGF cohort, central foveal thickness decreased significantly at one-month follow-up (p < 0.001).
DeepME showed substantial agreement with manual grading (p < 0.001).
Interpretation:
DeepME demonstrates high accuracy in evaluating anti-VEGF treatment response and shows potential for real-world clinical decision support.
Conclusion:
DeepME is a novel clinical decision support system that integrates an improved YOLOv11 detection architecture for comprehensive macular edema management.
