The ophthalmic disease, age-related macular degeneration (AMD), is one of the several eye diseases that afflict the health of thousands of people globally. Computational tools based on machine learning models can be viable alternatives aimed at assisting ophthalmologists in the process of diagnosing and understanding fundus images of macular degeneration. However, it is evident that access to advanced and refined ophthalmic medical mechanisms is a distant reality for many healthcare professionals. Therefore, the development of a method that uses simpler instruments and materials, such as fundus images, can be an alternative for these specialists who need diagnostic assistance but do not have the necessary materials. In light of this, this experimental work presents a study that combines backbones (feature extractors), tree-based classifiers, and PCA (Principal Component Analysis) to develop a binary image classification method. Using the ODIR-5K (Ocular Disease Intelligent Recognition) dataset, we obtained fundus images of Normal class and Age-Related Macular Degeneration class with the aim of developing a model that highlights the main characteristics of these classes and classifies them. Analyzing the results, we achieved 0.85 Precision and 0.96 Recall for the Normal class, and 0.88 Precision and 0.98 Specificity for the AMD class. Therefore, these results demonstrate the potential of the proposed experimental work in terms of binary classification of these two classes.