Abstract: Oral diseases such as dental caries and gingivitis represent a significant global public health problem, requiring accurate and accessible diagnostic methods. This work proposes an automated classification system based on a Convolutional Neural Network (CNN) to distinguish between these two pathologies from clinical images. The implemented architecture consists of four convolutional layers with Batch Normalization and Max-Pooling, followed by fully connected layers with regularization by Dropout, achieving an accuracy of 95.3% on the test set. Data augmentation techniques, including rotations, flips, and contrast adjustments, were employed to enhance model generalization, along with early stopping strategies and learning rate scheduling to optimize training. The results demonstrate an excellent balance between sensitivity (92% for gingivitis) and specificity (99% for caries), with an average F1-score of 0.95. This research evidences the potential of CNNs as tools to support dental diagnostics, while highlighting the importance of addressing challenges such as inter-patient variability and integration with existing clinical systems for future implementations in real-world settings.

Keywords: Caries, Classification, Deep Learning, Diagnosis, Gingivitis, Images, Neural

Pages: 133-139