Neurosymbolic AI Approach for Dry and Wet AMD Classification Using OCT Images

Accurate classification between dry and wet age-related macular degeneration (AMD) is crucial for determining appropriate treatment strategies and improving patient outcomes. However, deep learning convolutional neural networks (CNNs) often face challenges when working with small datasets, particularly in the context of rare pathologies, which can hinder their robustness and generalizability. To address these issues, we employed a neurosymbolic approach that integrates medical knowledge in the form of symbolic AI, enhancing the model’s interpretability and reasoning capabilities. The aim of this study was to improve the classification of retinal conditions, specifically dry AMD, wet AMD, and healthy retinas. Our results demonstrated an overall accuracy of 93%, indicating the effectiveness of this methodology in accurately classifying retinal diseases. These findings suggest that the neurosymbolic approach holds promise for advancing diagnostic support in ophthalmology while providing a transparent decision-making framework.