Improved Dental Disorder Diagnosis Using Hybrid EfficientNet-CoAtNet Feature Extraction and Stacking Classifiers

Accurate disease diagnosis with X-ray images is essential for prompt treatment and better patient outcomes in the dental field. In order to diagnose dental diseases from X-ray images, a new deep learning framework is proposed in this paper. A hybrid approach is used in our framework. The high-level semantic features captured by EfficientNet are combined with the feature importance refinement capabilities of CoAtNet. The result of this combination is a richer and more informative feature representation for disease classification. These features are then efficiently integrated using a stacking classifier to produce a reliable and broadly applicable model. Our framework's effectiveness is demonstrated by evaluating a dataset of X-ray images, where it outperforms other existing methods. Specifically, our model obtains an average accuracy equal to 93.8% accuracy, precision equal to 93.1%, sensitivity equal to 93.2%, specificity equal to 93.3%, and Dice similarity coefficient equal to 93%. These findings point to the potential of our framework as a valuable resource for dental practitioners, supporting precise and timely disease detection for the best possible patient care.