Applications of Explainable Artificial Intelligent Algorithms to Age-related Macular Degeneration Diagnosis: A Case Study Based on CNN, Attention, and CAM Mechanism

Abstract

The recent years have witnessed extensive research on Explainable Artificial Intelligence (XAI) algorithms in the field of ophthalmology. This paper introduces an improved deep learning model for Age-related Macular Degeneration (AMD) classification by incorporating XAI techniques such as Convolutional Neural Networks (CNN), Attention, and Class Activation Mapping (CAM). The study compares the model’s explainable indicators, accuracy, and test time (per image) with those of the original model. The results demonstrate that the utilization of interpretable algorithms in ophthalmology holds significant clinical potential, serving as an effective means to establish trust between artificial intelligence and the medical field. Moreover, it proves to be a viable approach to enhance accuracy. Therefore, this research contributes to the reference significance of XAI application in ophthalmology.