Recent advancements in deep learning have significantly improved medical imaging diagnostics, notably in identifying eye conditions such as glaucoma. However, the opaque nature of these models raises trust issues, especially in healthcare applications. To enhance the interpretability of such systems, this study integrates Explainable Artificial Intelligence (XAI) into deep learning-based glaucoma prediction. The integration employs LIME-driven super-pixel segmentation techniques, specifically SLIC, Quick Shift, and Felzenszwalb, to evaluate interpretability. Comparative analysis reveals that the Felzenszwalb method delivers superior performance in visual and statistical interpretability, making it most suitable for clinical insight.