Non-proliferative diabetic retinopathy detection using Rosmarus Quagga optimized explainable generative meta learning...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Non-Proliferative Diabetic Retinopathy (NPDR) is an early-stage complication of diabetes affecting retinal blood vessels.
• Existing diagnostic models struggle with lesion heterogeneity (size, shape, distribution).
• Accurate and early NPDR diagnosis is crucial for diabetes management.

Purpose of the Study:

• To develop and validate a novel model for effective diagnosis of severe diabetic retinopathy (DR) with lesion recognition.
• To address challenges in existing NPDR detection models.
• To improve the accuracy and interpretability of DR diagnosis using retinal images.

Main Methods:

• Proposed a Rosmarus Quagga optimized Explainable generative Meta learning based Deep Convolutional Neural Network (RQ-EGMCN).
• Integrated adaptive foraging and leader-based feeding strategies for enhanced detection accuracy.
• Employed explainable CNNs with attention/saliency maps for interpretability.
• Utilized generative components for realistic retinal image synthesis and meta-learning for accelerated generalization.

Main Results:

• The RQ-EGMCN model achieved high performance on a diabetic retinopathy detection dataset.
• Maximum accuracy of 95.47%, precision of 95.34%, and recall of 95.24% were recorded.
• The model demonstrated improved diagnostic accuracy, reduced computational complexity, and enhanced versatility.

Conclusions:

• The proposed RQ-EGMCN model offers an effective solution for NPDR diagnosis.
• Explainable AI components enhance decision-making transparency in DR detection.
• The model shows significant potential for clinical application in diabetic eye care.