Glaucoma Detection using Golden Jackal Optimization based MobileNet Dual Attention Classification

Early identification of glaucoma, world's biggest cause of irreversible blindness, is crucial for preventing vision loss; nevertheless, this is an arduous and time-consuming process when done manually. Using deep learning, this research suggests a way to automatically identify glaucoma in improved retinal scans taken from Kaggle dataset. Algorithm classifies retinal pictures for the purpose of diagnosing glaucoma and creates masks for optic disc and optic cup based on those analyses. Most up-to-date techniques for feature classification and image preprocessing are utilised in this study to provide a thorough approach to glaucoma detection. When preprocessing a dataset, Contrast Limited Adaptive Histogram Equalisation (CLAHE) is used to reduce noise and enhance dataset quality. By combining the best features of many MobileNet designs, the hybrid model boosts classification accuracy. For the purpose of glaucoma image classification, a MobileNet-based Dual Attention Mechanism (MobileNet-DualAttention) classifier has been used. To further enhance the accuracy of the MobileNet-based Dual Attention model, the parameters are hyperparameter tuned using the Golden Jackal Optimisation Algorithm (GJOA). With a test set accuracy of 99.86%, the system proved its better performance during validation on a Kaggle dataset. A generalisable and practical strategy for glaucoma detection, these findings demonstrate efficacy of suggested method.