Development and Validation of an Explainable Convolutional Neural Network Platform for Skeletal Malocclusion Classification and Patient Education with Lateral Cephalometric Radiographs
Skeletal malocclusion, Lateral cephalometric radiographs, Explainable artificial intelligence, Convolutional neural network, EfficientNet-B0, Grad-CAM, SHAP, Orthodontic diagnosis, Clinical decision support.
AuthorsAbstractSkeletal malocclusion diagnosis, particularly through lateral cephalometric x-rays, remains an important part of orthodontic methodology. Nevertheless, several AI techniques fail to be clinically intelligible despite their wellacknowledged performance in terms of classification accuracy. This study presents a novel approach to skeletal malocclusion diagnosis based on an explainable convolutional neural network (XAI-CNN) approach that includes elements of EfficientNet-B0 and Grad-CAM, while also relying on SHAP to classify malocclusions into Class I, II, or III categories with the aim of visual explanations. The proposed methodology encompasses a set of standardized image processing operations and techniques for data augmentation and transfer learning to enhance the overall system ability to attain reliable and accurate results. Furthermore, the patient education interface helps improve the communication process in terms of presentation AI-generated with explanations. Overall evaluation indicates that the proposed approach outperforms traditional CNN methods in terms of classification accuracy, effectiveness, and explainability.’’
•••••••••••••••••••••••••••••••• ejprd.org- Published by Riset Publishing Services LLC.
EJPRD
Copyright © 2026 by Riset Publishing Services LLC