A unified framework of information measures for complex linear Diophantine fuzzy sets with application to chronic kidney disease.

The inferior alveolar nerve (IAN), a branch of the mandibular branch of the trigeminal nerve, supplies sensation to the lower teeth and lip. This nerve is an important nerve in dentistry that is encountered when extracting mandibular third molars or replacing lost or extracted teeth with implants. The IAN may also be encountered during orthognathic surgery or surgeries for various cystic, benign, and malignant pathologies. Complications from procedure-related IAN damage cause most of the recent legal medical disputes in dentistry. Identifying the course/location of the IAN before treatment is clinically important in preventing nerve damage. The introduction of cone-beam computed tomography (CBCT) has made identifying the course of the IAN easier and more accurate than was possible using plain radiography. However, challenges remain for clinicians in identifying the intuitive three-dimensional (3D) course of the IAN. With the development of artificial intelligence (AI) and deep learning technologies, clinical applications, particularly in the field of medical imaging, are being actively performed. Learning techniques using convolutional neural networks (CNNs) can automatically perform object detection, specifying, classification, and segmentation in images. This technology enables automatic and intuitive segmentation of specific anatomical structures in images, and its accuracy is gradually increasing. Specifically for dentistry, the automatic detection and identification of the course of the IAN and the ability to dynamically track the nerve during oral procedures will decrease the oral surgeon’s burden, leading to faster and safer surgeries. Our previous study aimed to automatically segment the IAN by training CBCTs of 98 patients using a customized “no-new-net” (nnU-Net) based on a 3D U-net. Through three rounds of repeated learning, we observed a dice similarity coefficient (DSC) of approximately 0.58 ± 0.08 and a segmentation time of 86.4 Sect.. As learning was repeated, a gradual increase in accuracy and decrease in extraction time was observed; however, the final extracted results were limited in clinical application.