Dental Rugae Pattern Analysis for Craniofacial Growth Pattern Recognition Using InceptionV3: A Machine Learning Approach

Craniofacial growth pattern analysis is of paramount importance in accurate diagnosis and orthodontic treatment planning. Conventional cephalometric analysis is time consuming and subjects patients to several radiographic surveys. This paper examines the relationship between craniofacial growth patterns and palatal rugae, and presents a deep learning approach with the InceptionV3 architecture for automated growth pattern classification (vertical, average, and horizontal). Intraoral images collected from The Oxford Dental College and augmented by various transformations made up the dataset used to fine-tune the pre-trained model through transfer learning. The ensuing model attained a test accuracy of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{8 1. 1 3 \%}$</tex>, with confusion matrices and Grad-CAM visualizations presenting strong performance for horizontal and average patterns, but difficulties in accurate vertical pattern classification. These results present the possibility of deep learning in improving dental rugae analysis for orthodontic applications, while presenting areas of future improvements through improved datasets and model calibration.