Classification of Glaucoma in Retinal Images Using EfficientnetB4 Deep Learning Model

Today, many eye diseases jeopardize our everyday lives, such as Diabetic Retinopathy (DR), Age-related Macular Degeneration (AMD), and Glaucoma. Glaucoma is an incurable unavoidable disease that damages the vision of optic nerves quality life. Classification has been active field research for past ten years. Several approaches classification are established, beginning with conventional segmentation methods feature-extraction to deep-learning techniques Convolution Neural Networks (CNN). In contrast, CNN classifies input images directly using tuned parameters convolution pooling layers by extracting features. But, volume training datasets determines performance CNN; model trained small datasets, overfit issues arise. therefore developed transfer learning. The primary aim this study explore potential EfficientNet learning current work compares other models, namely VGG16, InceptionV3, Xception public RIM-ONEV2 & V3, ORIGA, DRISHTI-GS1, HRF, ACRIMA. dataset split into training, validation, testing ratio 70:15:15. assessment test shows pre-trained EfficientNetB4 achieved highest value compared models listed above. proposed method 99.38% accuracy also better results metrics, sensitivity, specificity, precision, F1_score, Kappa score, Area Under Curve (AUC) models.