This paper introduces a novel method for the automatic classification of common throat diseases, with a particular focus on mononucleosis. We first explore the role of artificial intelligence (AI) in enhancing oral disease diagnosis by reviewing successful classification techniques and demonstrating how the integration of both domain-specific and general oral cavity images can improve clinical decision-making. Recognizing the scarcity of high-quality, publicly available datasets, we detail our comprehensive data collection, preprocessing, and augmentation procedures, which resulted in a curated dataset of 832 images labeled with the assistance of the Victor Babeș University of Medicine and Pharmacy Timisoara students. Notably, 382 additional images were generated using a Generative Adversarial Neural Network (GANN) to address class imbalance. Furthermore, we describe the design and training of an ensemble model comprising three binary classifiers that achieve robust performance metrics of 90% accuracy, 91% precision, 90% recall, 93% ROC-AUC, and 90% F1 score. We conclude by comparing our results with the current state of the art, discussing limitations, and suggesting potential improvements in model architecture and dataset expansion.