AttnOpt-DentSegNet: An Attention-Enhanced and Optimized Deep Learning Framework for Panoramic Dental Image Segmentation

One of the most important steps when utilizing computer-aided oral health treatment planning and diagnosis assessments is efficiently segmenting dental structures in radiographs through panoramic images. However conventional models developed using deep learning algorithms, on the other hand, have significant limitations because of the considerable wide variety of structural features and the existence of medical artifacts. Here, in this research article, we present AttnOpt-DentSegNet which represents an intelligent framework to dental computational image analysis through optimization-guided segmentation with an emphasis on attention enhancement. Our architectural design makes use of spatial attention gates in skip connections, which are an extension of the deep U-Net model. This allows the deep neural network to focus on relevant regions and ignore irrelevant ones. By automatically tuning critical hyperparameters such as learning rate and filter count as well as dropout rate, we develop an adaptive particle swarm optimization (APSO) technique for substantially enhancing the generalization and training efficiency. In order to train and assess the performance of our proposed deep learning framework, we utilized a panoramic dental image dataset that contains 1,978 dental image-mask pairs. Applying several numerical and qualitative experiments has shown that AttnOpt-DentSegNet delivers greater performance in terms of different machine learning-based computer vision evaluation metrics such as intersection over union (IoU), dice coefficient, and accuracy when compared to baseline and state-of-the-art models, and also different optimization variations at the same level of performance. It has been demonstrated by our findings that the utilization of attention-driven architecture together with adaptive hyperparameter optimization leads in generating segmentation results that shows great robustness and highly accurate results in a complex area such as visual radiography in dentistry.