Classification Precision in Endodontic Imaging: Advanced Deep Learning with Adaptive Squeeze-and-Excitation in Enhanced VGG-19 and Feature Pyramid Network

Precisely classifying and detecting endodontic treatment obturation is critical for effective dental diagnostics and optimal oral health. Conventional strategies are time-consuming and often need more precision. This study presents a novel deep-learning-based architecture, VGG19Fpn-ASENet, which is a modification of the VGG-19 network incorporating a Feature Pyramid Network (FPN) and an Adaptive Squeeze-and-Excitation Network (ASENet) to enhance performance in this application. The existing SENet, known for its efficiency and compactness, has been refined by integrating these modifications to enhance feature extraction and adaptability. The proposed VGG19Fpn-ASENet model was evaluated using an open-access endodontic treatment image dataset, achieving 100% accuracy in both training and validation phases and 99.51 % accuracy on the test set. This significant advancement exhibits the model's prospect of greatly enhancing diagnostic accuracy and efficiency in dental care.