Evaluation of the relationship between maxillary molar tooth roots and maxillary sinus with artificial intelligence

Aim: The maxillary sinus is a critical structure that plays a significant role in dental procedures due to its close proximity to the maxillary teeth and its importance in daily dental practice, especially regarding issues like cysts and tumors. It has been reported that 10.5-34.2% of the roots of maxillary molar teeth are in close relation to the floor of the maxillary sinus. Various complications can occur as a result of procedures applied to maxillary molar teeth. The aim of our study is to determine the relationship between the roots of maxillary molar teeth and the maxillary sinus using artificial intelligence algorithms. Method: This retrospective study involved the creation of four different ROI (Region of Interest) areas from panoramic radiographs of 500 patients, resulting in a total of 7,000 ROIs. Classification was performed based on whether or not there was a relationship between the roots of the maxillary molar teeth and the maxillary sinus. Hybrid models that combine deep learning and machine learning, such as VGG16+SVM, VGG19+SVM, and InceptionV3+SVM, were used. Results: The highest accuracy was achieved with the ROI area of 312*357 pixels and the InceptionV3 model. For this ROI area and the InceptionV3 model, precision, recall, F1 score, and accuracy were 0.80, 0.81, 0.71, and 0.70, respectively. Conclusion: Successful results are obtained when the relationship between maxillary molar teeth and the maxillary sinus is examined using artificial intelligence algorithms. Detecting this relationship in panoramic radiographs, which are frequently used in clinical practice, may reduce the need for CBCT imaging, which has limited accessibility, and higher doses and costs compared to panoramic radiography.